AUTOMATION HISTORY

The Last Driver License Holder… …is already born. How Waymo, Tesla, Zoox & Co will change our automotive society and make mobility safer, more affordable and accessible in urban as well as rural areas. CES 2026 clearly showed that physical AI is coming in the form of autonomous cars and now also humanoid robots. At least 144 companies are developing humanoid robots, offering more than 170 models. Around 30 manufacturers were represented at CES 2026, along with just as many component suppliers who develop and offer robot hands, among other things. It is noteworthy how many companies from the automotive industry are involved in the development and use of humanoids. I am aware of at least six automotive companies (OEMs) that are developing complete humanoid robots. These and others, such as BMW and BYD, are using them in their own factories in production or as information representatives (Chery den AIMOGA). Some Tier 1 suppliers in the automotive industry are also collaborating with humanoid robot manufacturers on development, or have even acquired a robot manufacturer (see Mobileye, which recently acquired Mentee Robotics). Estimates by analyst firms such as McKinsey that one billion humanoid robots will populate the earth by 2040 and take over tasks in factories and households demonstrate the market potential. A trillion-dollar market is opening up here, which could become even larger than the automotive market. And that could also be a lifeline for automotive companies that are currently undergoing massive upheaval. And it makes sense: humanoid robots are not only a new form of mobility, but the theoretical capabilities required to develop and mass-produce them are very much in line with the current capabilities of automotive manufacturers and suppliers. At the same time, they would solve another problem: they could be used in the companies’ own production facilities, where they could perform their work more cheaply, without being unionized, and without high turnover and the increasingly urgent shortage of skilled workers due to monotonous and unhealthy tasks. In this respect, it appears that no automobile manufacturer or supplier can avoid dealing with humanoid robots and participating in their development. This article was also published in German. View all posts by Mario Herger Δ

Oops, something went wrong LAS VEGAS, Jan 6 (Reuters) - Mobileye Global on Tuesday said it would acquire humanoid robotics startup Mentee Robotics for about $900 million, as the ​Israeli self-driving technology firm bets on what it sees as the ‌next frontier of artificial intelligence. Mobileye stock rose 12% in early trading on Wednesday. The deal highlights the overlap between autonomous driving and robotics, where similar sensing, ‌perception and decision-making technologies underpin the emerging field of embodied AI. Interest in humanoid robotics, in particular, is surging, driven by the idea that human-like forms can better adapt to existing warehouses, factories and complex settings, helping to ease labor shortages and ⁠boost productivity. Intel spun out its ‌computer vision business RealSense last year to speed expansion into robotics. It also remains the largest shareholder in Mobileye with about ‍a 23% stake. Amnon Shashua, who serves as the CEO of Mobileye, cofounded Mentee Robotics and is the startup's co-CEO. Mentee raised about $21 million in a funding round in March, valuing ​the startup at roughly $162 million, according to PitchBook data. The company counts Cisco ‌and Samsung's VC arms among its investors. Tesla, Figure AI, Agility Robotics and several Chinese startups are among the companies racing to develop two-legged robots capable of performing a wide range of tasks. Tesla CEO Elon Musk expects humanoid robots to become the company's largest business in the long term. The deal, announced at the CES ⁠technology show in Las Vegas, brings together Mobileye's ​software, sensing and safety systems for self-driving ​cars with Mentee's development of general-purpose humanoid robots. Mentee Robotics says it bypasses the need for massive real-world data collection to train the ‍robot by transforming a ⁠single human demonstration into millions of virtual repetitions. First proof-of-concept deployments with customers are expected in 2026, with series production and commercialization targeted for 2028. It ⁠added that the transaction, subject to customary closing conditions, is expected to close in the first ‌quarter of 2026. (Reporting by Akash Sriram in Bengaluru and Abhirup Roy ‌in Las Vegas; Editing by Vijay Kishore) Sign in to access your portfolio

Themes Humanoid Robotics ETF (BOTT US) – Portfolio Construction Methodology The underlying Solactive Global Humanoid Robotics Index targets global all-cap equities exposed to humanoid, service and industrial robotics. The index universe comprises constituents of the Solactive GBS Developed Markets All Cap, South Korea All Cap and China All Cap indices with minimum 1- and 6-month ADVT of USD 1,000,000 and a single most liquid share class per issuer, subject to a 60% liquidity buffer for existing lines. Solactive’s ARTIS natural language processing identifies companies generating at least 50% of revenue from defined humanoid robotics value-chain segments, scores their thematic relevance and removes non-theme names. Securities are ranked by score; the top 6 always qualify, existing members ranked 7–36 are retained, and additional names are added until 30 constituents are reached. Constituents receive relevance-based weights derived from rank, then are capped at 4.5% each. The index is reconstituted and rebalanced quarterly on the first Wednesday of February, May, August and November. To explore BOTT in more depth, visit our ETF analytics platform for institutional-grade insights — including performance and risk metrics, correlations, sensitivities, and factor exposure: https://www.etfstrategy.com/etf/BOTT_US Comments are closed. Use of this website is subject to the terms of our disclaimer, cookies statement and privacy policy. By continuing to browse the site, you are indicating your acceptance of these terms.

Researchers at China’s Tsinghua University’s Shenzhen International Graduate School have developed a next-generation tactile sensor called SuperTac. The project involved collaboration with multiple domestic and international institutions and was inspired by the unique visual structure of pigeon eyes. The research aims to solve one of robotics’ biggest challenges: enabling robots to sense touch with human-like precision and understanding. The study was published on January 15 in Nature Sensors under the title “Biomimetic Multimodal Tactile Sensing Enables Human-like Robotic Perception.” As embodied intelligence advances, robots are moving beyond controlled factory floors into human-centered environments. This shift demands high-resolution tactile sensing, multimodal perception, and better interpretation of physical contact. Current tactile sensors still struggle with limited resolution and weak data fusion, making touch one of the main bottlenecks in robotic perception. SuperTac is a multimodal, high-resolution tactile sensor that combines multispectral imaging (from ultraviolet to mid-infrared) with triboelectric sensing signals. Its ultra-thin, multi-layer sensing skin allows micrometer-level resolution and can detect force, contact position, temperature, proximity, and vibration. The system can identify material type, texture, slippage, collision, and even color with over 94% accuracy. To process this complex tactile data, the team developed DOVE, an 850-million-parameter tactile language model. DOVE enables robots to interpret touch information in a more human-like way, significantly improving environmental understanding and manipulation accuracy. The SuperTac breakthrough also reflects the broader China–US humanoid AI competition. China is currently ahead in robot hardware, sensors, and large-scale deployment, supported by strong manufacturing and fast research-to-product cycles. The US, meanwhile, leads in AI software, foundation models, and autonomous intelligence, driven by companies like Tesla, Figure AI, and Boston Dynamics. In the short term, China may dominate physical humanoid deployment, while long-term leadership will depend on who best combines advanced hardware with powerful AI brains. SuperTac has already been integrated into robotic dexterous hands, enabling real-time tactile feedback. Looking ahead, this technology could transform manufacturing, medical robotics, and service robots, bringing the industry closer to robots that can truly see, think, and feel like humans. Read More: (via)

ByIan Sayson, Forbes Staff. SoftBank Group—controlled by Japan’s richest person, Masayoshi Son—has agreed to buy the robotics unit of Swiss industrial giant ABB Ltd. for $5.4 billion, deepening the Japanese company’s AI push. “SoftBank’s next frontier is physical AI,” Son, chairman and CEO of Softbank, said in a statement. “Together with ABB Robotics, we will unite world-class technology and talent under our shared vision to fuse artificial super intelligence and robotics—driving a groundbreaking evolution that will propel humanity forward.” The acquisition, expected to be completed by mid to late 2026, follows SoftBank’s set up earlier this year of Robo HD, a new holding company that owns its robotics assets. Over a dozen portfolio companies from Agile Robots SE and Skild AI were moved to Robo HD, which analysts say is an indication of Son’s renewed interest in robotics. Son’s earlier interest in robotics led to the creation of Pepper, a humanoid unveiled in 2014 but its production was halted later after demand waned. SoftBank is also invested in technologies aimed at advancing industrial automation and logistics. “SoftBank will be an excellent new home for the business and its employees,” ABB CEO Morten Wierod said in the statement. “ABB and SoftBank share the same perspective that the world is entering a new era of AI-based robotics and believe that the division and SoftBank’s robotics offering can best shape this era together.” ABB’s robotics business—which employs 7,000 people spread in manufacturing hubs in China, the US and Sweden—booked $313 million in EBITDA on $2.3 billion revenue in 2024, compared with $385 million and $2.5 billion in 2023. With a net worth of $67.4 billion based on Forbes real-time data, Son, 68, is pouring billions of dollars into AI, including his investment in the $500 billion Stargate AI infrastructure project in the U.S., in partnership with OpenAI, Oracle and MGX.

From daily news and career tips to monthly insights on AI, sustainability, software, and more—pick what matters and get it in your inbox. Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies. We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. Follow Us On Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. The hybrid leg uses a five-bar linkage with 12 motors, placing most near the pelvis to cut leg mass and improve dynamic walking models. Researchers in the US developed bipedal robots with a new design, the HybridLeg platform, to advance reinforcement learning.Featuring a lantern-shaped, sensorized mechanical cover, these robots can safely handle whole-body contact. To tackle the inherent instability of humanoids, Univeristy of Illinois’ Kinetic Intelligent Machine LAB (KIMLAB) developed a protective design that mitigates fall impacts and allows autonomous recovery to a standing posture, enabling self-reset after each trial. Combined with multimodal fall detection and enhanced stance phase tracking, this platform paves the way for robust, long-horizon, real-world reinforcement learning experiments. A video, shared by KIMLAB, details an innovative untethered bipedal robot featuring a unique “hybrid leg” mechanism. While traditional humanoid robots typically use serial linkages to mimic human anatomy, this design combines the biological familiarity of serial linkages with the mechanical advantages of parallel linkages, such as higher speeds, lower inertia, and superior payload-to-weight ratios. The Hybridleg is a parallel linkage mechanism in which each link consists of a serial chain, forming a five-bar closed linkage. The robot utilizes 12 motors for actuation, with a significant design choice to concentrate 10 of them near the pelvis, leaving only 2 at the ankles. This configuration drastically reduces distal mass, which minimizes the negative impact of swing leg dynamics and allows for more accurate physics modeling using reduced-order models like the linear inverted pendulum As a fully self-contained, untethered platform, the robot houses all necessary components—including a single-board computer, IMU, voltage converter, and LiPo batteries—within its body. The presentation concludes by demonstrating the robot’s capabilities through various walking experiments. The large-scale bipedal robot demonstrates how a hybrid mechanical design can push humanoid locomotion toward greater agility, strength, and efficiency. Detailed in a recent paper, the robot is built around the HybridLeg mechanism, a novel approach that combines serial and parallel structures to deliver six degrees of freedom per leg while maintaining low inertia and a large workspace. The design enables faster motion, higher payload capacity, and improved dynamic performance—key requirements for agile bipedal walking. To further enhance structural rigidity and precision, the latest version of the HybridLeg is fabricated using carbon fiber tubes and high-precision bearings, allowing the structure to support its own weight without sacrificing accuracy. A pair of HybridLegs is assembled into a full bipedal platform using a custom pelvis design inspired by human biomechanics. The pelvis incorporates a yaw angle offset, similar to the toe-out angle in human feet, to expand the reachable workspace of the feet and improve overall stability. Simulation results detailing workspace and velocity ranges are validated through hardware experiments, confirming close agreement between theory and practice. The robot stands 1.84 meters tall—taller than the average human—while weighing just 29 kilograms (64 pounds). Despite its size, it can be driven by the same class of servo motors typically used in smaller humanoid robots, highlighting the efficiency of the hybrid mechanism and optimized structural design. The paper provides a detailed explanation of the mechanical architecture, along with full kinematic analysis and analytical solutions. Performance validation encompasses multi-body dynamics simulations, as well as preliminary hardware experiments, including squatting and in-place walking motions. According to researchers, a simple forward walking demonstration further confirms the feasibility of the approach. Together, these results position the Hybrid Leg-based biped as a promising platform for future research in humanoid locomotion, scalable robot design, and real-world dynamic walking experiments. Jijo is an automotive and business journalist based in India. Armed with a BA in History (Honors) from St. Stephen's College, Delhi University, and a PG diploma in Journalism from the Indian Institute of Mass Communication, Delhi, he has worked for news agencies, national newspapers, and automotive magazines. In his spare time, he likes to go off-roading, engage in political discourse, travel, and teach languages. Premium Follow

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript. Advertisement Scientific Reports , Article number: (2026) Cite this article 554 Accesses 1 Altmetric Metrics details We are providing an unedited version of this manuscript to give early access to its findings. Before final publication, the manuscript will undergo further editing. Please note there may be errors present which affect the content, and all legal disclaimers apply. The field of child-robot interaction (CRI) is growing rapidly, in part due to demand to provide sustained, personalized support for children in educational contexts. The present study uses a within-subject design to compare how children between 5 and 8 years of age (n=32) interact with a robot and human instructor during a tangram learning task. To assess how the children’s characteristics may influence their behaviors with the instructors, we correlated interaction metrics, such as eye gaze, social referencing, and vocalizations, with parent-reported scales of children’s temperament, social skills, and prior technology exposure. We found that children gazed more at the robot instructor and had more instances of social referencing toward a research assistant in the room while interacting with the robot. Age was related to task time completion, but few other individual characteristics were related to behavioral characteristics with the human and robot instructors. When asked about preferences and perceptions of the instructors after completing the tangram tasks, children showed a strong preference for interacting with the robot. These findings have implications for the integration of social technologies into educational contexts and suggest individual differences play a key role in understanding how children will uniquely respond to robots. Please contact the corresponding author, Allison Langer, at Allison.langer@temple.edu for access to de-identified data files. Belpaeme, T. et al. Guidelines for designing social robots as second language tutors. Int. J. Soc. Robot. 10, 325–341 (2018). Google Scholar Johal, W. Research trends in social robots for learning. Curr. Robot. Rep. 1(3), 75–83 (2020). Google Scholar Johal, W., Castellano, G., Tanaka, F. & Okita, S. Robots for learning. Int. J. Soc. Robot. 10, 293–294 (2018). Google Scholar Van den Berghe, R., Verhagen, J., Oudgenoeg-Paz, O., Van der Ven, S. & Leseman, P. Social robots for language learning: A review. Rev. Educ. Res. 89(2), 259–295 (2019). Google Scholar Belpaeme, T., Kennedy, J., Ramachandran, A., Scassellati, B. & Tanaka, F. Social robots for education: A review. Sci. Robot. 3(21), 5954 (2018). Google Scholar Conti, Daniela, Cirasa, C., Di Nuovo, S. & Di Nuovo, A. Robot tell me a tale! A social robot as tool for teachers in kindergarten. Interact. Stud. 21(2), 220–242 (2020). Google Scholar Fong, Frankie TK., Sommer, Kristyn, Redshaw, Jonathan, Kang, Jemima & Nielsen, Mark. The man and the machine: Do children learn from and transmit tool-use knowledge acquired from a robot in ways that are comparable to a human model?. J. Exp. Child Psychol. 208, 105148 (2021). Google Scholar Papadopoulos, I. et al. A systematic review of the literature regarding socially assistive robots in pre-tertiary education. Comput. Educ. 155, 103924 (2020). Google Scholar Zhexenova, Z. et al. A comparison of social robot to tablet and teacher in a new script learning context. Front. Robot. AI. 7, 99 (2020). Google Scholar Serholt, S., Ekström, S., Küster, D., Ljungblad, S. & Pareto, L. Comparing a robot tutee to a human tutee in a learning-by-teaching scenario with children. Front. Robot. AI. 9, 836462 (2022). Google Scholar Tolksdorf, N.F., Mertens, U. “Beyond words: Children’s multimodal responses during word learning with a social robot”. In International Perspectives on Digital Media and Early Literacy. Routledge 90-102 (2020) Admoni, H. & Scassellati, B. Social eye gaze in human-robot interaction: A review. J. Human-Robot Interact. 6(1), 25–63 (2017). Google Scholar Serholt, Sofia, Barendregt, W. Robots tutoring children: Longitudinal evaluation of social engagement in child-robot interaction. In Proc. 9th Nordic Conference on Human-Computer Interaction. (2016). Mwangi, Eunice, N. Gaze-based interaction for effective tutoring with social robots. (2020). Sim, Gavin & Bond, R. Eye tracking in child computer interaction: Challenges and opportunities. Int. J. Child-Comput. Interact. 30, 100345 (2021). Google Scholar Walle, Eric A., Reschke, P. J. & Knothe, J. M. Social referencing: Defining and delineating a basic process of emotion. Emot. Rev. 9(3), 245–252 (2017). Google Scholar Tolksdorf, N.F., Crawshaw, C.E., Rohlfing, K.J. Comparing the effects of a different social partner (social robot vs. human) on children’s social referencing in interaction. Front. Educ. Front. Med. SA 5 (2021). Tolksdorf, Nils F., Viertel, F. E. & Rohlfing, K. J. Do shy preschoolers interact differently when learning language with a social robot? An analysis of interactional behavior and word learning. Front. Robot. AI 8, 676123 (2021). Google Scholar Di Dio, C. et al. Shall i trust you? From child–robot interaction to trusting relationships. Front. Psychol. 11, 469 (2020). Google Scholar Baxter, P., De Jong, C., Aarts, R., de Haas, M., Vogt, P. The effect of age on engagement in preschoolers’ child-robot interactions. In Proc. Companion of the 2017 ACM/IEEE Int. Conf. Human-Robot Interact. (2017). Van Straten, C. L., Peter, J. & Kühne, R. Child–robot relationship formation: A narrative review of empirical research. Int. J. Soc. Robot. 12(2), 325–344 (2020). Google Scholar Neumann, M. M., Koch, L. C., Zagami, J., Reilly, D. & Neumann, D. L. Preschool children’s engagement with a social robot compared to a human instructor. Early Child. Res. Quart. 65, 332–341 (2023). Google Scholar Wilson J. R, Aung P.T, Boucher I. “When to help? A multimodal architecture for recognizing when a user needs help from a social robot”. In International Conference on Social Robotics. Cham (Springer Nature. Switzerland, 2022) Barry, Ruby-Jane., Neumann, M. M. & Neumann, D. L. Individual differences and young children’s engagement with a social robot. Comput. Hum. Behav.: Artif. Hum. 4, 100139 (2025). Google Scholar Kanero, J. et al. Are tutor robots for everyone? The influence of attitudes, anxiety, and personality on robot-led language learning. Int. J. Soc. Robot. 14(2), 297–312 (2022). Google Scholar Nasvytienė, Dalia & Lazdauskas, T. Temperament and academic achievement in children: A meta-analysis. Eur. J. Investig. Health Psychol. Educ. 11(3), 736–757 (2021). Google Scholar Raptopoulou, Anastasia, Komnidis, A., Bamidis, P. D. & Astaras, A. Human–robot interaction for social skill development in children with ASD: A literature review. Healthc. Technol. Lett. 8(4), 90–96 (2021). Google Scholar Flanagan, T., Wong, G. & Kushnir, T. The minds of machines: Children’s beliefs about the experiences, thoughts, and morals of familiar interactive technologies. Dev. Psychol. 59(6), 1017 (2023). Google Scholar Andries, V. & Robertson, J. Alexa doesn’t have that many feelings: Children’s understanding of AI through interactions with smart speakers in their homes. Comput. Educ.: Artif. Intell. 5, 100176 (2023). Google Scholar Mott, T., Bejarano, A., Williams, T. Robot co-design can help us engage child stakeholders in ethical reflection. In 2022 17th ACM/IEEE International Conference on Human-Robot Interaction (HRI). (IEEE, 2022). Rubegni, E., Malinverni, L., Yip, J. Don’t let the robots walk our dogs, but it’s ok for them to do our homework: Children’s perceptions, fears, and hopes in social robots. In Proc. 21st Annual ACM Interaction Design and Children Conference (2022). de Jong, C., Peter, J., Kühne, R. & Barco, A. Children’s intention to adopt social robots: A model of its distal and proximal predictors. Int. J. Soc. Robot. 14(4), 875–891 (2022). Google Scholar Leyzberg, D., Spaulding, S., Toneva, M., Scassellati, B. The physical presence of a robot tutor increases cognitive learning gains. In Proc. Annual Meeting of the Cognitive Science Society 34 34 (2012). Kirschner, D., Velik, R., Yahyanejad, S., Brandstötter, M., Hofbaur, M. YuMi, come and play with Me! A collaborative robot for piecing together a tangram puzzle." Interactive Collaborative Robotics. In 1st International Conference, ICR 2016, Budapest, Hungary, Proceedings 1. Springer International Publishing 24-26 (2016). Resing, W. C., Vogelaar, B. & Elliott, J. G. Children’s solving of ‘Tower of Hanoi’tasks: Dynamic testing with the help of a robot. Educ. Psychol. 40(9), 1136–1163 (2020). Google Scholar Zinina, A., Zaidelman, L., Kotov, A., Arinkin, N. The perception of robot’s emotional gestures and speech by children solving a spatial puzzle. Computational Linguistics and Intellectual Technologies. In Proc. International Conference. Dialogue 19 26 (2020). Yang, Yuqi, Langer, A., Lauren Howard, L., Peter J. Marshall P.J, Wilson J.R. Towards an ontology for generating behaviors for socially assistive robots helping young children. In Proc. AAAI Symposium Series 2(1) 213-218 (2023). Rothbart, M. K., Ahadi, S. A., Hershey, K. L. & Fisher, P. Investigations of temperament at three to seven years: The children’s behavior questionnaire. Child. Dev. 72(5), 1394–1408 (2001). Google Scholar Gresham, F., Elliott, S. N. Social skills improvement system (SSIS) rating scales. Bloomington, MN: Pearson Assessments (2008). Gülay Ogelman, H., Güngör, H., Körükçü, Ö. & Erten Sarkaya, H. Examination of the relationship between technology use of 5–6 year-old children and their social skills and social status. Early Child Dev. Care 188(2), 168–182 (2018). Google Scholar O’Reilly, Z., Roselli, C., Wykowska, A. Does exposure to technological knowledge modulate the adoption of the intentional stance towards humanoid robots in children? (2023). Kennedy, James, Baxter, Paul & Belpaeme, Tony. The impact of robot tutor nonverbal social behavior on child learning. Front. ICT. 4, 6 (2017). Google Scholar Nicolas, Sommet, Weissman, D. L., Cheutin, N. & Elliot, A. J. How many participants do I need to test an interaction? Conducting an appropriate power analysis and achieving sufficient power to detect an interaction. Adv. Method. Pract. Psychol. Sci. 6(3), 25152459231178730 (2023). Google Scholar van Straten, C. L., Peter, J. & Kühne, R. Transparent robots: How children perceive and relate to a social robot that acknowledges its lack of human psychological capacities and machine status. Int. J. Hum.-Comput. Stud. 177, 103063 (2023). Google Scholar Chernyak, N., & Gary, H. E. Children’s cognitive and behavioral reactions to an autonomous versus controlled social robot dog. In Young Children’s Developing Understanding of the Biological World 73-90 (Routledge, 2019). Kahn, P. H. Jr. et al. Robovie, you’ll have to go into the closet now: Children’s social and moral relationships with a humanoid robot. Dev. Psychol. 48(2), 303 (2012). Google Scholar Chen, Y. C., Yeh, S. L., Lin, W., Yueh, H. P. & Fu, L. C. The effects of social presence and familiarity on children-robot interactions. Sens. (Basel). 23(9), 4231 (2023). Google Scholar Almousa, Ohoud & Alghowinem, Sharifa. Conceptualization and development of an autonomous and personalized early literacy content and robot tutor behavior for preschool children. User Model. User-Adapt. Interact. 33(2), 261–291 (2023). Google Scholar Maaz, Nafisa, Mounsef, J. & Maalouf, N. CARE: Towards customized assistive robot-based education. Front. Robot. AI 12, 1474741 (2025). Google Scholar Langer, A., Marshall, P. J. & Levy-Tzedek, S. Ethical considerations in child-robot interactions. Neurosci. Biobehav. Rev. 151, 105230 (2023). Google Scholar Download references We would like to thank all the children and their families for participating in this project. A special thanks to our research assistants: Emily Peeks and Georgia May for their hard work on numerous aspects of this project including data collection and coding, Chelsea Rao for designing the robot behaviors and programming the interface, and Yuqi Yang for operating the robot and her assistance in data collection.This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. 2038235 and via the Franklin & Marshall Hackman Summer Research Scholars Fund. Temple University, Philadelphia, USA Allison Langer & Peter J. Marshall Franklin & Marshall College, Lancaster, USA Jason R. Wilson & Lauren Howard Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Allison Langer wrote the manuscript and prepared all the figures. All authors contributed substantially to reviewing and editing the manuscript and were involved in the study design. Lauren Howard led data collection and behavioral coding efforts, and Jason Wilson instructed all robot-related programming, extracted gaze data, and supplied the team with the Misty Robot. Allison performed data analyses in R. Peter Marshall provided compensation for the families in the study from research funds. Correspondence to Allison Langer. The authors declare no competing interests. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/. Reprints and permissions Langer, A., Wilson, J.R., Howard, L. et al. The influence of individual characteristics on children’s learning with a social robot versus a human. Sci Rep (2026). https://doi.org/10.1038/s41598-025-27476-x Download citation Received: 30 August 2024 Accepted: 04 November 2025 Published: 14 January 2026 DOI: https://doi.org/10.1038/s41598-025-27476-x Anyone you share the following link with will be able to read this content: Sorry, a shareable link is not currently available for this article. Provided by the Springer Nature SharedIt content-sharing initiative Advertisement Scientific Reports (Sci Rep) ISSN 2045-2322 (online) © 2026 Springer Nature Limited Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.

From daily news and career tips to monthly insights on AI, sustainability, software, and more—pick what matters and get it in your inbox. Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies. We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. Follow Us On Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. A robot dog learns adaptive walking on rough terrain using deep reinforcement learning trained fully in simulation. A quadruped robot has learned to walk across slippery, uneven terrain entirely through simulation, without any human-designed gaits or manual tuning. The system relies on deep reinforcement learning to adapt its movement in unpredictable environments. Legged robots are widely seen as useful tools for disaster response, exploration, and industrial inspection. However, traditional control methods struggle when surfaces change suddenly or terrain becomes unstable. These systems depend on precise physical models and predefined motion patterns, which often fail outside controlled settings. Deep reinforcement learning offers an alternative by allowing robots to learn from experience. But training instability and poor performance on unfamiliar terrain have limited its real-world use. Many learned controllers perform well only in the environments they were trained in. The study addresses these limitations by introducing a structured learning framework. Instead of exposing the robot to complex environments from the start, the system trains it gradually using a curriculum that increases terrain difficulty step by step. The quadruped robot learns to walk efficiently and maintain balance without any human intervention, showing strong adaptability to surfaces it has never encountered before. The robot is modeled with 12 degrees of freedom and controlled using a hierarchical structure. A high-level neural network policy runs at 10 Hz and generates target joint movements. These commands are executed by a low-level proportional-derivative controller running at 100 Hz to ensure stable and accurate motion. To understand its surroundings, the robot combines internal sensing with simulated vision. Proprioceptive inputs include joint angles, velocities, and body orientation. Exteroceptive data comes from a simulated depth camera that provides local terrain heightmaps, slope estimates, and friction information. Training uses the proximal policy optimization algorithm. The reward function balances several objectives, including forward speed, stability, smooth motion, low energy use, and reduced foot slippage. This multi-objective setup encourages natural walking behavior rather than rigid or inefficient movement. A four-stage curriculum plays a central role. Training begins on flat ground, then progresses to slopes, rough terrain, low-friction surfaces, and finally mixed environments with added sensor noise. This gradual increase allows the robot to build robust locomotion skills. Testing was carried out in the Webots simulator across multiple terrain types. The trained controller achieved forward speeds between 0.79 and 0.9 meters per second while maintaining low energy consumption and minimal slippage. Fall rates ranged from 0 percent on flat ground to 12 percent on low-friction terrain. The policy also generalized well to new environments, achieving a 94.6 percent success rate in Webots and 91.2 percent in the PyBullet simulator without retraining. Ablation studies showed that curriculum learning was critical. Models trained without it experienced higher fall rates and increased energy usage. In contrast, the full framework reduced falls to 5 percent and slippage to 4.2 percent. The robot also developed emergent behaviors during training. These included shifting weight laterally on slopes, shortening strides on rough ground, and stepping cautiously on slippery surfaces. None of these behaviors were explicitly programmed. Despite strong simulation results, the researchers note challenges in transferring the system to real-world robots. Hardware limitations, sensor inaccuracies, and environmental unpredictability remain obstacles. Future work will focus on reducing the sim-to-real gap using techniques like domain randomization and hybrid control systems. The study demonstrates that adaptive legged locomotion can emerge entirely from simulation, bringing autonomous robots closer to real-world deployment.The study appears in the journal Scientific Reports. With over a decade-long career in journalism, Neetika Walter has worked with The Economic Times, ANI, and Hindustan Times, covering politics, business, technology, and the clean energy sector. Passionate about contemporary culture, books, poetry, and storytelling, she brings depth and insight to her writing. When she isn’t chasing stories, she’s likely lost in a book or enjoying the company of her dogs. Premium Follow

@2022 - All Right Reserved. Designed and Developed by PenciDesign Davanti a tre contenitori di magliette di taglie differenti, Figure 03 non esita. Il robot individua quella richiesta e la porge con precisione. Un gesto semplice, eppure capace di mostrare ciò che oggi distingue i dispositivi umanoidi più avanzati: l’integrazione tra visione, linguaggio e azione. Il robot della startup californiana Figure AI è stato progettato per compiti pratici in ambienti controllati. E il suo ultimo modello segna un passo significativo rispetto al passato. Il robot è alto circa 1,68 metri, con rivestimenti in tessuto e imbottiture in schiuma per limitare i rischi negli incontri con gli esseri umani. Il dispositivo è più leggero del 9% rispetto ai predecessori, dotato di altoparlanti più potenti e di ricarica wireless nei piedi. Ciò con un’autonomia di circa cinque ore. Ogni funzione, dalla voce alla manipolazione degli oggetti, è pensata per rendere più fluide le interazioni. Anche se le risposte vocali richiedono ancora due-tre secondi, interrompendo la naturalezza del dialogo. Il segreto delle sue capacità è Helix, il modello Vision-Language-Action di Figure AI, che collega ciò che il robot percepisce, comprende e compie. Tale sistema consente a Figure 03 di passare rapidamente dalla comprensione di una richiesta verbale alla sua esecuzione pratica. Come nel caso della consegna delle magliette. Presentato ufficialmente nell’ottobre 2025, Figure 03 si propone come assistente robotico per attività di pick-and-place, con applicazioni in contesti domestici e professionali. Le innovazioni introdotte rispetto ai modelli precedenti mostrano come i robot umanoidi stiano riducendo il divario tra capacità tecniche e interazione con l’uomo. Il video social che ha immortalato la dimostrazione ha acceso discussioni e curiosità. Evidenziando, in tal modo, tanto le potenzialità quanto i limiti di un robot destinato a diventare sempre più familiare negli ambienti quotidiani. Figure 03 non è solo tecnologia: è un’anticipazione di come gli umanoidi potrebbero entrare nelle nostre vite quotidiane nei prossimi anni. Ciao sono Margareth, per gli amici Maggie, la vostra amichevole web writer di quartiere. Questa piccola citazione dice già tanto di me: amo il cinema, le serie tv, leggere e cantare a squarciagola i musical a teatro. Se a questo aggiungiamo la passione per la fotografia e la tecnologia direi che è facile intuire perché ho deciso di studiare e poi lavorare con la comunicazione. 2012 – 2023 Tecnoandroid.it – Gestito dalla STARGATE SRLS – P.Iva: 15525681001 Testata telematica quotidiana registrata al Tribunale di Roma CON DECRETO N° 225/2015, editore STARGATE SRLS. Tutti i marchi riportati appartengono ai legittimi proprietari. Questo articolo potrebbe includere collegamenti affiliati: eventuali acquisti o ordini realizzati attraverso questi link contribuiranno a fornire una commissione al nostro sito.

Figure AI enchaîne les démonstrations pour prouver l’agilité de son robot humanoïde. Cette fois-ci, son fondateur partage une vidéo de footing avec des employés. Grâce à leur capacité à automatiser une variété de tâches, en usine et à la maison, les robots humanoïdes font partie des marchés les plus prometteurs du secteur de la tech. Et parmi les sociétés qui pourraient devenir leader de ce marché, il y a Figure. Régulièrement, Brett Adcock, fondateur de cette startup, publie des vidéos qui mettent en avant les progrès de son robot humanoïde. Et, récemment, celui-ci a partagé une vidéo qui montre l’agilité de ce robot pour la course à pied. “Annonce d’un nouveau programme de remise en forme chez Figure”, a-t-il écrit dans sa publication, dont la vidéo montre un groupe d’employés faire un footing avec le robot. Announcing new fitness program at Figure pic.twitter.com/0PcGljahL1 — Brett Adcock (@adcock_brett) January 15, 2026 Tout en gagnant en agilité, le robot humanoïde de Figure devient aussi de plus en plus intelligent. Celui-ci a fait parler de lui en 2025 grâce à des vidéos sur lesquelles on voit le robot exécuter de nouvelles tâches ménagères qu’il a apprises. Et, récemment, Brett Adcock a également présenté un système de recharge par induction qui permet au robot de se recharger juste en étant debout sur une plateforme. Figure 03 is capable of wireless inductive charging Charging coils in the robot’s feet allow it to simply step onto a wireless stand and charge at 2 kW In a home setting, this means the robot can automatically dock and recharge itself as needed throughout the day There was… pic.twitter.com/xigMEhKg85 — Brett Adcock (@adcock_brett) January 8, 2026 Cette année pourrait être décisive pour le développement des robots humanoïdes. Dans une autre publication, le fondateur de Figure prédit qu’en 2026, “les robots humanoïdes effectueront des tâches non supervisées, sur plusieurs jours, dans des maisons qu’ils n’ont jamais vues auparavant, entièrement pilotés par des réseaux neuronaux.” Pour rappel, l’un des concurrents de Figure, 1X, prévoit de commercialiser son robot Neo dès cette année. Alors que la plupart de ses concurrents se concentrent sur les applications industrielles, 1X a fait le choix de cibler les usages domestiques. 4 predictions in 2026: 1. Humanoid robots will perform unsupervised, multi-day tasks in homes they’ve never seen before – driven entirely by neural networks. These tasks will span long time horizons, going straight from pixels to torques 2. Electric vertical takeoff and landing… — Brett Adcock (@adcock_brett) January 1, 2026 Pour le moment, les robots humanoïdes sont un tout petit marché. D’après les chiffres d’Omdia, relayés par Forbes, 13 317 unités auraient été “expédiées” en 2025. Cependant, ces expéditions pourraient presque doubler, chaque année, pour atteindre les 2,6 millions d’unités en 2035. 📍 Pour ne manquer aucune actualité de Presse-citron, suivez-nous sur Google Actualités et WhatsApp. Newsletter 🍋 Abonnez-vous, et recevez chaque matin un résumé de l’actu tech J'ai lu et accepte les termes et les conditions états-unishumanoïde [ Source ] Sur le même sujet Comment le fils de Donald Trump dédie sa vie à parier sur l’actualité Pourquoi les sous-marins russes sont-ils les seuls à avoir été conçus en titane ? Historique : Wall Street se modernise, la Bourse va tourner 24/7 sans ouverture ni clôture Elon Musk dégaine déjà son chéquier pour influencer les prochaines élections américaines ILLIMITÉ 20 Go 5,99 € ILLIMITÉ 350 Go 19,99 € Mode Suivez-nous Mode Suivez-nous

ByJohn Koetsier, Senior Contributor. Airbus is dipping its toes into the humanoid robot workforce with a purchase of China-based UBTech’s robots, according to Bloomberg. The European airplane manufacturer is apparently going to try to put humanoid robots on the assembly line, as the deal includes a plan to work together with UBTech on integrating robots into the airplane-building process. Airbus ordered UBTech’s Walker S2, a full-size humanoid that stands 176 cm tall (5’9"), weighs 70 kg (154 lbs), and walks at about two meters/second (4.5 mph). It has dextrous hands with 11 degrees of freedom and tactile sensors, and can hold 7.5 kg (16.5 lbs) in each hand and 1 kg (2.2 lbs) with each finger. The Walker S2 can also hot-swap its own batteries, and UBTech said in November of 2025 it was the first humanoid robot to be able to do so. Particularly useful for production environments, the Walker S2 can pivot on its waist almost 180 degrees, enabling it to quickly move components or work on other parts without shifting its feet. There is an external emergency stop button and power switch, located on the robot’s back. UBTech has already shipped about 1,000 humanoid robots, putting it in third place behind Agibot and Unitree for shipments globally and ahead of western robotics manufacturers like Figure AI, Agility Robotics, Tesla, Apptronik, and Hyudai-owned Boston Dynamics. It’s also ahead of European humanoid robot companies like Neura Robotics, based in Berlin, and the newly-funded Generative Bionics, which is based in Italy. UBTech is targeting a production capacity of 5,000 industrial humanoid robots this year, and 10,000 for 2027, and as of November 2025 said that it had already sold $112 million worth of humanoid robots. If the reported production numbers are correct, that would indicate an average six-figure price tag for each humanoid: $112,000 each. That price should fall quickly as UBTech scales production over the next few years. The companies did not announce details of the agreement, including the number of robots purchased or pricing. Almost certainly this is an initial deal for demonstration and testing purposes. I’ve asked Airbus if the company is also buying and testing humanoid robots from European manufacturers, and will update this story as the company responds. The Bank of America has estimated that mass adoption of humanoid robots will start in 2028. Because it’s so significant to labor costs and manufacturing capability, humanoid robot development is the “space race of our time,” says Apptronik CEO Jeff Cardenas, which has raised hundreds of millions of dollars. If so, China is half-way to the moon, and other countries have some catching up to do.

Experience See the campus, culture, and people that make Yale Engineering a top-ranked program. Experience Overview > Study Explore our engineering disciplines and degree programs that shape tomorrow's innovators. Study Overview > Research / Faculty Discover our strategic Areas of Impact where pioneering faculty are driving breakthroughs in today's most critical fields. Research / Faculty Overview > News / Events Excitement pulses through every corner of our campus daily. Stay connected with the latest engineering breakthroughs, stories and events. News / Events Overview > About Learn about our leadership, vision and commitment to excellence in engineering education. About Overview > When it comes to training robots to perform agile, single-task motor skills, such as hand-stands or backflips, artificial intelligence methods can be very useful. But if you want to train your robot to perform multiple tasks—say, performing backward flip into a hand-stand—things get a little more complicated. “We often want to train our robots to learn new skills by compounding existing skills with one another,” said Ian Abraham, assistant professor of mechanical engineering. “Unfortunately, AI models trained to allow robots to perform complex skills across many tasks tend to have worse performance than training on an individual task.” A dog-like robot from Abraham’s lab executes a controlled flip, demonstrating how hybrid control theory enables smooth transitions between learned motor skills. Think of how we learn new skills or play a sport. We first try to understand and predict how our body moves, then eventually movement becomes muscle memory and so we need to think less. To solve for that, Abraham’s lab is using techniques from optimal control—that is, taking a mathematical approach to help robots perform movements in the most efficient and optimal way possible. In particular, they’re employing hybrid control theory, which involves deciding when an autonomous system should switch between control modes to solve a task. One use of hybrid control theory, Abraham said, is to combine the different methods in which robots learn new skills. These include, for instance, methods of learning from experience with reinforcement learning, or through model-based learning in which robots plan their actions through observing. The trick is getting the robot to switch between these modes in the most effective way so that it can perform high-precision movements without losing performance quality. “Think of how we learn new skills or play a sport,” he said. “We first try to understand and predict how our body moves, then eventually movement becomes muscle memory and so we need to think less.” Abraham and his research team applied methods from hybrid control theory to a dog-like robot, training it to successfully balance and then flip over As part of the robot training, AI methods are used to develop the more challenging motor skills that require whole-body precision. Hybrid control theory is then leveraged to schedule and synthesize the various mechanisms for robots to learn diverse motors skills to gain more complex, compounding behaviors. Ideally, this could lead to robots working in homes and other unstructured environments. “If a robot needs to learn a new skill when deployed, it can draw from its arsenal of learning modality, using some level of planning and reasoning to ensure safety and its success in the environment, with on-the-job experience,” Abraham said. “Once a certain level of confidence is achieved, the robot can then use the more specialized learned skills to go above and beyond.” Nov 20, 2025 Computer Science Mechanical Engineering Artificial Intelligence Robotics for Humanity AI, rat whiskers, and robots in the wild: Talking with Ian Abraham Yale Engineering's top 10 stories of 2025 Copyright © 2026 Yale University. All rights reserved.Website designed and developed by WORX.

El CEO de Nvidia, Jensen Huang, ha expresado públicamente su entusiasmo por una compañía de inteligencia artificial que está revolucionando la industria de la robótica: Serve Robotics. Esta empresa, dedicada al desarrollo de robots autónomos de entrega, ha captado la atención no solo del líder de Nvidia, sino también de importantes analistas de Wall Street que proyectan un crecimiento explosivo para sus acciones durante 2026, con estimaciones que sugieren un incremento potencial del 77%. La admiración de Jensen Huang hacia Serve Robotics no es casual. La compañía ha demostrado capacidades tecnológicas sobresalientes en el desarrollo de sistemas de navegación autónoma y algoritmos de inteligencia artificial aplicados a la logística de última milla. Cabe destacar que Nvidia mantiene una relación estratégica con diversos fabricantes de robótica, proporcionando sus procesadores especializados para aplicaciones de IA, lo que convierte el respaldo de Huang en un indicador relevante sobre el potencial de la empresa. Los analistas de Wall Street han identificado una oportunidad de mercado valorada en aproximadamente $450 mil millones de dólares para el sector de robótica de entrega. Las acciones de Serve Robotics han experimentado movimientos significativos en meses recientes, con incrementos superiores al 25% en diversas jornadas durante diciembre de 2025 y enero de 2026. Sin embargo, este comportamiento también refleja la volatilidad característica de empresas emergentes en sectores tecnológicos de alto crecimiento. El interés institucional ha aumentado considerablemente, especialmente tras anuncios de expansión operativa y nuevas asociaciones estratégicas que fortalecen su posición competitiva en el mercado de entregas autónomas. A decir verdad, la proyección del 77% de crecimiento para 2026 representa una apuesta ambiciosa pero fundamentada en el potencial disruptivo de la tecnología de Serve Robotics. Sin lugar a dudas, el respaldo explícito de figuras como Jensen Huang añade credibilidad al proyecto, aunque los inversionistas deben considerar los riesgos inherentes a empresas en etapas tempranas de desarrollo. La combinación de inteligencia artificial avanzada, robótica autónoma y un mercado en expansión podría convertir a esta compañía en uno de los jugadores más relevantes del sector durante los próximos años. Fuente: Yahoo Utilizamos cookies para ofrecerte la mejor experiencia en nuestra web. No compartimos tu información con nadie, simplemente queremos que navegues normalmente sin tener que rellenar formularios en cada visita. Puedes aprender más sobre qué cookies utilizamos o desactivarlas en los ajustes. Esta web utiliza cookies para que podamos ofrecerte la mejor experiencia de usuario posible. La información de las cookies se almacena en tu navegador y realiza funciones tales como reconocerte cuando vuelves a nuestra web o ayudarnos a comprender qué secciones de la web encuentras más interesantes y útiles. Nunca almacenamos información personal. Tienes toda la información sobre privacidad, derechos legales y cookies en nuestra página de privacidad y cookies. Las cookies estrictamente necesarias tiene que activarse siempre para que podamos guardar tus preferencias de ajustes de cookies. Esta web utiliza Google Analytics para recopilar información anónima tal como el número de visitantes del sitio, o las páginas más populares. Dejar esta cookie activa nos permite mejorar nuestra web.

Wisdom And Wonder Equanimity Λ Computing Costanza asked Sussman why MIT had switched away from Scheme for their introductory programming course, 6.001. This was a gem. He said that the reason that happened was because engineering in 1980 was not what it was in the mid-90s or in 2000. In 1980, good programmers spent a lot of time thinking, and then produced spare code that they thought should work. Code ran close to the metal, even Scheme — it was understandable all the way down. Like a resistor, where you could read the bands and know the power rating and the tolerance and the resistance and V=IR and that’s all there was to know. 6.001 had been conceived to teach engineers how to take small parts that they understood entirely and use simple techniques to compose them into larger things that do what you want. But programming now isn’t so much like that, said Sussman. Nowadays you muck around with incomprehensible or nonexistent man pages for software you don’t know who wrote. You have to do basic science on your libraries to see how they work, trying out different inputs and seeing how the code reacts. This is a fundamentally different job, and it needed a different course. So the good thing about the new 6.001 was that it was robot-centered — you had to program a little robot to move around. And robots are not like resistors, behaving according to ideal functions. Wheels slip, the environment changes, etc — you have to build in robustness to the system, in a different way than the one SICP discusses. And why Python, then? Well, said Sussman, it probably just had a library already implemented for the robotics interface, that was all. (via wingolog) So the reason, basically, is that software today is a train wreck, and you might as well embrace helplessness and random tinkering from the start? Horrifying. Unbelievable. I don’t know about 6.001, but if is a introductory course for non software engineers / CSers, I totally agree: practical reasons… This is a shame. I’ve been reading their introductory programming course, because I felt that using C or something higher (mostly VB in my case) is quick, but does not teach you how to be what I call a ‘real programmer’, and it doesn’t. Reading their Scheme-based course is an eye-opener. I see the reasoning behind their move to Python, which may have more immediate application and even make more sense to someone who does not have a background in CS, but really, it will only further the financial interests of MIT. I can’t help but feel someone Higher Up (duh duh duuuhhhhh) has pushed this on them. MIT will now churn out countless applications programmers. Half decent ones I am sure, but not the kind of full-on nerds that make the *very best* programmers. There seems to be such a divorce between systems and hard mathematical programmers on one hand, and applications programmers on the other. But regardless, programmers who have been raised on really, really geeky stuff are the best ones. EOF. 😉 I started with Pascal and BASIC, moved on to 80×86 assembly and *then* went to C and C++. I spent my school years breaking the school systems and finding out how they worked. These days I would say that I am a very bad programmer (though lack of practice), but I am still a million times better than someone who has been taught using only Java or Python, because I have a basic but functional knowledge of PCs and a feel for the workings of hardware and limitations of systems. Going back to Scheme and Python: When I started reading their course using Scheme, and I saw the way it worked, it was like I had come to the top of a mountainous pass and suddenly saw the blue-green vista of endless, verdant fields stretching away before me. That is a feeling I have not had in the last 10 years with regards programming and it is a crying shame that forthcoming MIT first-years will not have the same degree of understanding. It’s not about doing things, it’s about being able to do them. Anyone who knows how can write a brilliant programme, but if you are not apprised of The Way, then you will *have* to settle for mediocrity. Python is better at teaching introductory and deeper knowledge. Scheme is a must know but Python is practical, but attracts many smart programmers because of the ability to prototype quickly and understand the problem at hand not the mountain of ramp up that scheme can do. I agree with this and look at it this way, it could have been worse, it could have been Java. Python is by far my favorite language I can use to prototype and developer real world applications from desktop, web, system admin, services, game development and many other things. Plus CS grads will be familiar still with lambdas, tuples, lists as they are used in CS, dictionaries and they will start out with a clean dictated style. Python I believe will continue to win out over Java definitely but more and more scientific and academic support. Pathetic. Now even MIT is pandering to the “But I won’t ever use language X on a REAL job crowd”. This is ridiculous and I expect better from an institution like MIT. Seriously, our programs need to be teaching students concepts first (Data structures, Information Architecture, Systems Thinking, How to judge code quality based on merits – like ease to produce, maintain, understand, performance, etc) and how to express those concepts using the various families of languages out there later. Choose the language that expresses the concepts you’re going to teach with the least noise. Since I’m partial to starting with data and data structures, Scheme & the elegance through which it expresses the concept of recursion is my pick. “You have to do basic science on your libraries to see how they work, trying out different inputs and seeing how the code reacts” This is a bizarre argument. Empirical “science” testing of your libraries is more or less required in a dynamic scripting language like Python, where the type system is extremely poorly defined, no compiler is available, and the closest substitute is vast unit test coverage. Michael Lang: did you read the blognpost that you are responding to? Your claim of why MIT switched language is at odds with the professor’s claim. Do you intend to call him a liar? Kip Mailer: standard Scheme also has a dynamic type system. The trouble with python as the first language people see is that it is inherently confusing. There are three different kinds of things that “do stuff”, not just functions, but also operators and statements (and lambdas, but that’s not taught at first). Parentheses are badly confusing to beginners. Consider: >>> max(3,5) 5 >>> print(3,5) (3, 5) >>> print 3,5 3 5 >>> max 3,5 Syntax error: invalid syntax # because print is a statement, whereas max is a builtin function. Huh? >>> a = () >>> print a () >>> a = (3,5) >>> print a (3,5) >>> a + (7,9) (3,5,7,9) >>> a + (7) TypeError: can only concatenate tuple (not “int”) to tuple # Because that’s arithmetic parens, not tuple parens >>> a + (7,) (3,5,7) Is it any wonder people who learn python first come to be afraid of parentheses? get over yourself Walkingbeard. Programming languages should never be taught in class. There is no engineering in programming languages. Familiarity with a Programming Languages can be achieved by reading online manual or a book. Concepts like Data Structures, System’s thinking are ideal for class room teaching. You need a language or how do you create and manipulate these? How do you learn how to make them actually work? You just inform students ahead of time about what language will be used. And get to real business on lectures. You just use it, students will teach themselves about the syntax and other details. I took 6.001 at MIT. I also took 6.099, the experimental python-version of 6.001 when they needed guinea pigs to fool around with their new robots. The politics behind the switch were nasty, you wont ever get good answers from a faculty member about this ( observe sussman’s dark wit above). That said, the new course was slaved over by folks like Ableson and Kaebling who did their best to produce an adequate successor to scheme – 6.001. the mostly likely one liner explanation for the switch: authority insists a change from scheme ; python supports anonymous lambda functions – sussman appeased; python is well supported; python is interpreted…which means a lesser mind might confuse it with scheme; teaching c as an introductory language would destroy that special moment every MIT student gets to experience when his first boss reviews those first 1000 lines of c he committed and starts to say: “let me show you something they didnt teach you in your C course” … and the newbie responds … “hah, they dont teach that at MIT. Im just really fucking hardcore” Gregory, no offence but… “Parentheses are badly confusing to beginners. Consider:” exhibit a: ))))))))))))) The thing about Python tuples is that simply having a , makes it a tuple. Same with unicode just u’. It takes away much of the verboseness of Java. It also is much more simple than C. Again, allowing the user to focus on the problem at hand making the language a tool not another obstacle. The beauty of encountering Scheme in 6.001 was that it broadened your mind and changed your thinking, like it or not. This is the hallmark of great education, no matter what the field. Not everyone realized or appreciated it at the time; it took years for me to have my epiphany. I doubt it would ever have happened without my experiences with Scheme. I’m sure that Abelson and co. are trying their hardest to do with Python what they did easily with Scheme. I have serious doubts that they can provide the same kind of educational experiences that they used to. The students are probably bright enough to overcome it eventually, but something of real value has been lost. JLR, While I agree whole heartedly, I think it is worth mentioning that the fundamental ideas behind lisp based programming are preserved in Sussman’s two classes, Computational Mechanics and Adventures in Symbolic Programming (6.945). Most course 6ers are proactive enough to seek out classes like these to delve deeper into worthy topics. Thrusting SICP material onto uninterested students is all weve really lost with python-6.001. Weve gained a pretty comprehensive and practical introduction to systems engineering. Kids who care can still get to the original material in other courses. Please excuse the “get off my lawn” moment, but in my day, Course 6-3 (CS) was all about computer science, and 6.001 provided the groundwork for that. Software engineering was relegated to one class, 6.170. “Engineering” was something that course 6-1 (EE) did, and if we had to sit through 6.002 and God help us 6.003, then they could suck it up and deal with 6.001. Computer science was the whole idea, and I don’t think people should have to be in the position of having to seek it out. That being said, I am actually fairly ignorant of the current state of affairs, so I could be overstating things. hah…alright I concede! Of course I was (am?) one of those cop out 6-2ers, which makes me something of a not-a-real computer scientist and not-a-real hardware engineer. I tool on fpgas now….go figure. for the record, I liked 6.099 when I took it. Although the original version devoted about 1/3 of the course to 6.003 material. I’ll admit that programming robots does sound like fun. Of course, if I were to do it now, I’d use Haskell 🙂 One thing I loved about learning 6.001 in Scheme was that it easily and unconsciously twisted my mind into a purely functional way of thinking. We went six weeks without being taught the “set” command, so it was impossible to change the value of a variable. It’s amazing how much we could do with purely functional programming, and 6.001 taught us to think of the cleaner, functional approach *first*. Today I code mostly in C++, and it’s sad that without unnamed lambda expressions and closures, I program by constantly manipulating state. But in reality, most modern programming languages focus on manipulating state, so perhaps the 6.001 approach was quaint but outdated? Python has generators, lambdas, closures, and purely run-time types, so despite the enormous syntax difference, it actually has a lot in common with Scheme. It’s surely a better choice than Java or C++. When Sussman talks about robustness, I think this is what he means: http://groups.csail.mit.edu/mac/users/gjs/6.945/readings/robust-systems.pdf Maybe when Clojure is more mature, they’ll switch back to a Lisp. Scheme was great for folks who had programmed before. When I took 6.001, it was the first language I’d ever seen. Bad times. Took me years to learn to actually do anything. I had to laugh out loud at the “Python parentheses are confusing” comment. If that’s not Lisp=based satire, it’s damn close. Between learning CS or programming concepts/skills and a practical language, universities always have to strike a balance. Too much conceptual stuff, and you get research material, but not good for the industry; too much practical stuffs, and you become a vocational institute where your grads are one-trick ponies. > Parentheses are badly confusing to beginners. Consider:?)))))))))))))” Yes, this seems to be what non-lispers have trouble with until they realize just how helpful those parens are. Good tool support, e.g. paredit, is enough to overcome this. > The thing about Python tuples is that simply having a , makes it a tuple. It would be a might easier to teach if that were true, i.e. if this weren’t the case: >>> a = (,) SyntaxError: invalid syntax > The thing about Python tuples is that simply having a , makes it a tuple. Same with unicode just u?. It takes away much of the verboseness of Java. It also is much more simple than C. Again, allowing the user to focus on the problem at hand making the language a tool not another obstacle. Yes, this is what makes python worth criticizing. > Python has generators, lambdas, closures, and purely run-time types, so despite the enormous syntax difference, it actually has a lot in common with Scheme. I’m curious, could someone show me how to write (with-timeout 10 (lambda () foo) (lambda () timeout-handler)) using generators? They seem closer to iterators, but I’m a python newbie myself and may not have understood generators well enough. lambdas are one-liners only, are not very clearly delimited from their surroundings, and if I understand correctly will be getting removed from the language. closures are problematic: def make_bisector(initial_low,initial_high): low = initial_low high = initial_high mid = mean(low,high) def bisector(too_low): if too_low: low = mid else: high = mid mid = mean(low,high) return mid return bisector The = operator is used for both let and set!, so python here makes the decision that the = in the if implies a let at the level of the bisector function, rather than at the level of make_bisector. What one does in python is to use either a hash or an object to store low, mid, and high, but I think it’s hard for python to claim that it has useful closures here. The enormous syntax difference makes a difference. I’m currently teaching computing to students who have never programmed, and of necessity one must have some language to do it in, and MIT 6.00 uses python too, despite no robots. One of the biggest confusions is about return values vs. side effects (like printing), of course, but another was just between different kinds of things returning, and the order that things happen. I mentioned this before, but the syntax really gets in your way here. Operators take arguments to either side and return a value. Functions are on the left, and they use parens. Statements are on the left and they abhor parens (or rather, treat them as tuple parens, as in my print example). I have watched over and over my students’ eyes light up in understanding when I take the python syntax for a = 3 + max(5,8) and transform it to =(a, +(3, max(5,8))). They are not bothered by the “)))”. > I had to laugh out loud at the ?Python parentheses are confusing? comment. If that?s not Lisp=based satire, it?s damn close. I’m delighted to entertain. Grem Sorry, meaningful whitespace doesn’t always do well with indentation: def make_bisector(initial_low,initial_high): ~~low = initial_low ~~high = initial_high ~~mid = mean(low,high) ~~def bisector(too_low): ~~~~if too_low: ~~~~~~low = mid ~~~~else: ~~~~~~high = mid ~~~~mid = mean(low,high) ~~~~return mid ~~return bisector b = make_bisector(0,32) print b(True) print b(False) print b(True) Gives a syntax error: mid used before it is assigned. Ryan: “… it could have been worse, it could have been Java.” It is worse. 6.01 and 6.02 don’t teach all that much of what was in 6.001. The remainder of that has in theory been put in 6.005, “Principles of Software Development” … which uses Java. To paraphrase Wolfgang Pauli, for this purpose Java is not even wrong. I’ll note that this removes a “service subject” from the EECS department. A lot of people outside the department who had no interest in EE and who only wanted to learn computer science (as opposed to just programming) used to take 6.001, and now there is nothing for them in the department’s offerings (e.g. 6.005 has 6.01 as a prerequisite and 6.042J Mathematics for Computer Science as a corequisite). As for Sussman’s comment “…it probably just had a library already implemented for the robotics interface…”, well, no probably about it, it indeed uses just such a library. The politics behind all this are perhaps illustrated by the fact they have finished the purge of Scheme/LISP from the basic curriculum by switching the AI course (6.034) to Python. (I knew they were going to do this and a quick look at the course’s web site indicates it’s done; BTW, does anyone (good) in the real world do AI in Python???). Ah well. Gregory: Python 3.0 introduces the ‘nonlocal’ keyword to solve that problem. See here for more information: http://www.python.org/dev/peps/pep-3104/ You can also change the definition of bisector to the following to have it work: def bisector(too_low, mid=mid, low=low, high=high): I had several MIT grads in the developer training course I used to teach for a large software development company. They were bright, but the practicality (or lack thereof) of their programming education often left them far behind their classmates. I congratulate them for the soaring vistas they enjoyed when they first groked Scheme. So did I, when I learned it. But back here on planet Earth, we have software to develop to make the world a better place, and it’s not being done in Scheme. i appricate MITs dicision Gregory: I don’t think this issue is related to whitespace. Steven’s solution will remove the error, but will not give you what you want. You’re fighting Python’s object-oriented style. I think you’d want something like: http://pastebin.com/m2a457acf class Bisector: def mean(self,a,b): return (a+b)/2 def __init__(self,initial_low,initial_high): self.low = initial_low self.high = initial_high self.mid = self.mean(self.low,self.high) return def __call__(self,too_low): if too_low: self.low = self.mid else: self.high = self.mid self.mid = self.mean(self.low,self.high) return self.mid b = Bisector(0,32) print b(True) print b(False) print b(True) The only computer language ever giving me an hard on where SuperBasic on a Sinclair QL. Normal basic never had that appeal to me peek , poke whatever but it was much easier to do without looking through an LIBRARY to add some functionality to your code. Python ? well it works for today the learning curve is easy for beginners the fact of knowing what you want to import like any other language with the need of including something is annoying making the learn curve steep again. Going up a mountain could be hard , Decent from it even harder but the trip could be worthwhile. I second Schuyler’s Bisect class, except for the use of __call__. Not only are you pretending to be a simple function, you’re also returning a result (implying you have no side effect). As this particular case seems to need the side effect and the result, the next best thing is to find a method name that implies them both. Very good decision! All the difficulty this thread is having getting the bisector defined suggests that we still have a wee way to go. Who’d have imagined it more than 50 years after the invention of Fortran and Lisp… RIP 6.001 i’m from National University of Singapore, my lecturer was from MIT scheme class. And the teaching staff actually used scheme to create a library for us to program robots. We didn’t have to change to python If you don’t learn C, how in the H#ll can you be a CS magor? C is the mother tongue of all modern application languages, and YES you will learn how to use pointers. Dale Wilbanks: Yes, well, that’s the point of Joel Spolsky’s essay “The Perils of JavaSchools”, which says among other things that you really need to grok pointers and recursion. “If you don’t learn C, how in the H#ll can you be a CS magor?” — I was a 6-3/CS major at MIT and they never taught us C and I’ve done just fine. Languages you can pick up yourself by reading a manual. It’s the fundamental concepts behind all languages that they’re better off teaching. Unbelievable. I think software is now more of a crash site, not of our aircraft but of alien origin. You have so many things lying around but not quite really sure how to use them effectively until you dedicate your lifetime. Having actually taken, and aced, the now departed 6.001, I can say something meaningful about the course. It wasn’t about scheme, it was about what you could do with it. The more advanced software engineering course, 6.034 used something called CLU. I doubt anybody’s ever heard of it, but who cares. Computer science courses should teach fundamentals, and they probably should vary the language every few years if only to emphasize that it is primarily the approach to solving problems that really matters. Of course, that does NOT mean I want to go back to F77. But I could if I had to. Nobody ever accused MIT of producing bad programmers, even if they used languages that were not commercially prevalent. It’s a real shame that “.001” is gone. Actually, what I wrote isn’t strictly true. There was this whole debate about the “MIT approach” vs. the “NJ approach”, but those who know this also know there are merits to both. BTW, even when they taught 6.001, there were still courses available in Fortran and C. It is an error to say that C is not taught at MIT. I believe it was part of the Civil Engineering Dept at the time. I’m not going to look it up. I’m too lazy 🙂 This debate is silly. It is the “should we offer Latin since it is a dead language” debate. Fact of the matter is that you can learn about programming using any language. It’s the teaching and problem solving that’s important. Ubu Walker: I don’t think you understand what MIT is trying to accomplish with its EECS majors or the time constraints in a university course (13 weeks of instruction in a semester for MIT). In those contexts the choice of a language is very important. If you want to teach some functional programming you’d best not choose something like Java. You don’t want to spend a lot of time teaching syntax and solving syntax problems, so languages like Scheme and Python are good. I was saddened at first, as a Lisp hacker. But even in “my day”, most students complained about using Scheme the whole time. Only a fraction “got it”. Anyway, the 6.001 crew will be inspirational in Python. I doubt the students will lose much. The ones that get Lisp will still learn it. Many will be inspired to do so by the course. It is too bad to not be able to use SICP, as that was a truly great book. But times change. Well, students complained about every other course as well. I’m not sure where you get the idea that only a few people got it. The drop rate would have been a great deal higher were that true. Yet by the time of the eighth lab or so, most people were still in the course and attending lectures and recitations, so I would dispute the claim that \most people didn’t get it\ based on simple survival analysis. @Ubu: I disagree about “any language”. The language can get in the way – and I doubt machine code is a good language for this. A functional language has the most fundamentally rich concepts, and is appropriate at somewhere like MIT. Python is better than some, but it obscures some fundamental concerns. (And both obscure some fundamental concerns by being dynamically typed – concepts have important static structure in the real world.) Michael, sorry, I didn’t mean that students didn’t get 6.001, they did. I mean most students didn’t get the Lisp bug. But the purpose of the course isn’t to give us the Lisp bug, but to teach us programming, which most students got. Talk about a time lag – picking up a conversation seven months later. I still think that most got it – even the lisp side. Changing sides for a second, I think there were two places where people “got off the bus”. The first was at the mention of lambda. A fair # of People really were confused about how to properly employ anonymous functions. But based on % of people asking at recitation sections, maybe only 25% had this issue, and they had sorted it out. A larger number left (in spirit) during the metacircular evaluator assignment. Why take the time to write lisp in lisp? We already have it! It was a fairly long assignment that came at a time when all the other courses were getting tough as well. All said though, I still think dropping Scheme was a loss and a mistake. I really felt it served our goals well to think about programming rather than syntax. And the syntax was simple – it was just functions, and we “understand” functions well from our math classes. “Times” don’t change. People make decisions. there aren’t bad “times”, just bad cultures or bad institutions, or bad groupthink. “Times” didn’t change; we as a society decided to abandon an approach to programming where scheme/lisp would have excelled. I don’t like blaming “times” because it is a very bad figure of speech; it lets us blame circumstance instead of investigating the causes of group decision making. Ridiculous. Reading this, you’d think that scheme was a prerequisite to good programming. Come off it. The world is full of examples of excellent software, written by excellent programmers, with a myriad of backgrounds, many who’ve never written a line of scheme. Changing from scheme to python will make little or no difference to the outcomes of this course. The talented will be just as taltented regardless of the playground. If talent were the only issue, I’d have no problem with what jobbly is saying. First of all, Python came along well after scheme and absorbed some of its good ideas. So, it would not be surprising that we can obtain the same results. However, it is not simply identification of programming talent that is the objectives of this course. Scheme met the needs of a computer course well because one could get to the computer science ideas quickly without being mired down in programming details. It’s not simply a matter of “getting the job done” or “identifying talent”. A first year computer science course needs to meet the intellectual needs of the class, too. In other words, it needs to have a certain amount of elegance too. Now, I have not seen python so I cannot comment if this is elegant, but I reject your notion that all languages are equivalent for a first year computer science course. They could still do it in Fortran, and your logic would hold just fine. But they don’t, and for a good reason. Scheme was my first programming language in my Intro to CompSci class at the U. of Chicago in 1993. I was a second year undergrad at the time, and I had 0 experience with programming prior to that, and had learned to use email and Word a few months before. Our text was Abelson and Sussman. Needless to say…despite being pretty mathematically inclined, it kicked my ass. And the use of Scheme as the learning medium/tool did not help matters. The class was tough enough, the problem sets grueling and in some cases mind blowingly hard, but Scheme with its’ bare bones interface and joke level debugging did not help matters. I often spent more time being a good code technician than actually trying to solve the core problems being posed (8 queens problem anyone?) Python is an excellent teaching language. The interactive prompt is a nice tool for the students to use for experimentation and for me to demonstrate programming constructs in class. Python teaches good formatting habits. When my students use a free-format language and I tell them to observe good formatting conventions, they are trained to expect good formatting and they just do it. Python’s syntax is simple. It allows my students to focus on semantics instead of syntax. Python offers the additional benefit that is it a powerful tool and an excellent general-purpose programming language. I applaud MIT’s decision. We have been using Python since 2004 at NCSSM. It is interesting to see lots of others climbing aboard. What book does M.I.T use to teach python to their first year’s? We began using Python as a first programming language at the North Carolina School of Science and Mathematics in 2004. Our first course is a course in procedural programming, which uses Python objects as smart data types. Here are some benefits we have seen 1. Python’s grammar is simple. Hello world is print “Hello, World” or print (“Hello, World”) in Python 3. The students are not left bewildered by traffic jams of arcane keyworks. Python’s whitespace delimitation causes beginners to develop good formatting habits that last. There are not great piles of parentheses to wade through There are powerful libraries that do hosts of useful stuff. It supports the functional, imperative and OO paradigms but forces none of them on you. We like it a great deal. All introductory languages are intuitive at the start and have their nice features. It’s how easily you put the elements together that determines the true sense of a language. Syntactic sugar is sweet but fattening. I worry about non-issues like how to space things or matching parentheses. That’s my text editor’s job. I think succumbing to popular fads was a bad idea. I think a great deal of the assignments were made easier because we could start by solving problems without having to spend time on the syntax of the language. That said, Python was probably the best choice, given the regrettable decision to leave Scheme. I don’t agree. Coming with heavy assembly, C, C++ background Scheme was a train crash for me. I think Scheme is a beautiful tiny language with 1 page language spec, yet powerful enough to do even OOP. If you teach it right it is the best language you can teach to an average Joe and makes him a “very good” programmer in any language he might encounter in the future. Within a few weeks your average Joe starts developing his own algorithms within 2 months he shuffles objects around, in 6 months starts writing his own compiler. This is way faster learning curve than C/C++. Yet once you master it you become a very good programmer, the rest is just about learning the sytax rules of any language they propose you. I still do C/C++ for living (mostly on realtime) but even now for mocking up a small algorithm I use Scheme. Python inherits a lot from Scheme except “the consistent” but a bit hard-on-the-eyes syntax of Scheme. Consistency is important for a starter because with his one page spec knowledge he can decode any expression he might encounter, without diving deep in to the thick language manual (e.g. ANSI C spec and any later C/C++ specs). One good thing on Python’s account perhaps is that it has lots of open source support libraries so for anything which requires interaction with the “real world” (including GUI, Web, custom hardware) is easier. I think I would teach Scheme first to teach about programming, then switch to Python (which is relatively easy), for real world then C/C++/Java is just something you pickup when you need it. MIT’s switch from Scheme/Lisp to Python was a surprising move, agreed, but it and the discussion above demonstrates that computer science and the programming languages that implement them (and are of course themselves manifestations of the latest in CS, or at least attempt to be) are a alive and well, and, half a century on, are growing rather than slowing their pace towards maturity. So let professionals and amateurs alike rejoice and enjoy working or dabbling in the field, in spite of the mess. In fact, challenges and opportunities arise from the mess. Or would anyone rather be an expert designer of steam locomotives or build analog electronic circuits around pentodes? The physics of steam and vacuum tubes and the engineering disciplines of of their machines and circuits are well settled, no surprises there. In other words, all maturity and no mess. What a shame they did not pick Ruby, but as someone here already pointed out, this is pushed by someone higher up. So many people in their ivory towers looking down on the world in idealistic bliss, drowning in the depth of their pool of ego. None have ever strayed near the real world with it’s deadlines and profit margins. Oh no. They make their own toys to play with. The other thing people forget about 6.001 is that it’s a course for engineers, taught by engineers. They simply used the best tool for the job. Now, 20 odd years later, new tools have been made and they are better than the old tools. It’s called progress! And now ‘industry’ is moving towards functional programming, which is exactly what you learn in SICP. You can also write apps in Racket/Scheme and deploy directly to Yandex cloud. 6.001 is still the best intro to comp science Thanks for the auspicious writeup. It in reality was once a enjoyment account it. Glance complex to more introduced agreeable from you! By the way, how can we communicate? SICP is back at MIT in the form of 6.037 which teaches a “condensed” version. http://web.mit.edu/alexmv/6.037/ You can actually develop professional programs in Scheme using Lambda Native. You write out the guts in Scheme then can automatically port to iOS, Blackberry, Android, OSX, Windows, BSD, Linux ect. Write one program, export many times. More and more industry languages are becoming Lisp like, such as Julia. I think in the end all of us will just be using some kind of optimized Lisp. I’m glad Knuth didn’t take MIXAL/MMIX out of The Art of Programming books. He uses the same defense claiming what would be the point of including the low level flavor of the day when he’d have to go back and constantly rewrite his books to update the language. He also once had to learn a half dozen different low level languages to build compilers back in the 60s/70s so if he can be fluent in a number of different languages any computer scientist can. You learn the fundamentals in one, then just pickup a “learn Python (or Assembler)” book and translate what you know already in a weekend, learning new syntax and libraries. @SchemeHax, SICP isn’t really “back” at MIT — note that course is during IAP (January winter vacation during which offbeat and experimental courses are offered, some led by fellow students, and aren’t part of any degree program or necessarily overseen by academic departments). Source: am MIT Alumnus python programming language is best programming language.It is very useful for career.most of the company using the python program.Python Training Institutes in Chennai To start, python has 0 value for CS. For intro to programming python/java/csharp/etc. are all equally horrific but can do the job. However, there are two languages every CS must have at least rudimentary knowledge of: schema/lisp and c/c++. Lisp was just few weeks as part of intro to AI. Wonder how my thinking&approach about programming could have been different granted I was introduced to scheme instead of C first !!! My 10 year old wants to learn Programming. Scheme or Python – which is it? Simple English please. Python for sure. Makes math super easy too if you get into it. the ideas of programing in SCIP is permanant. I should probably write my own blog post on the topic, but, in short — When I took computer science courses, in a school to the south of MIT not known traditionally for engineering, the core two courses were “programming” (taught in M68000 assembler) and “algorithms and data structures” (taught in pascal). That’s where you were supposed to learn “all the way down”. Interestingly, most students “didn’t get it”, especially algorithms and data structures, because they could copy from the book. They didn’t re-think the problem. I asked for – and was allowed – to submit all my problem sets in C, which means I had to re-code everything and use only the ideas in the book, not the code itself. Thus, I submit, that while the old world of understanding “all the way down” has greatly vanished, the idea of skating over the top of unknown libraries has been with us since the beginning. I didn’t get an A in that course, and I did become my TA’s worst nightmare because I quickly knew more C than they did, but I came out with an extraordinary knowledge of data structures that form the core of my professional practice to this day. There were no other programming courses. I remember the first day in Operating Systems. Tom said “the course is taught in C. Anyone not know C? If you don’t know C, go to (some particular) TA, they will help you get up to speed and be your resource, but if you know programming you can program in any language. Let’s begin Operating Systems.” It’s funny you mention Data Structures and Algorithms, because in High School I took a summer class at Harvard (it’s a school a little northwest of MIT) that was also taught in Pascal, then I went on to get a CS degree from MIT, but I always felt most of my day-to-day programming knowledge came from that one Data Structures class. Arrays, pointers, structures, sorting and searching, binary trees, heaps and stacks were never really taught in the MIT comp sci program, which was mostly about theory and handling complexity, but it’s the stuff you use every day as a programmer. Dear fellows, Out of the lambda-calculus/lisp/scheme way I can’t imagine how I would have be able to build a language allowing me to easily explore basic concepts like lambdas, booleans, pairs, lists, recursion,… and lots of funny algorithms. Certainly not following the Python way. I wonder what you think about such a project: http://lambdaway.free.fr/lambdawalks/?view=fromroots2canopy http://lambdaway.free.fr/lambdaspeech Alain Marty Your email address will not be published. Required fields are marked * Comment * Name * Email * Website Notify me of followup comments via e-mail. You can also subscribe without commenting. Δ

From daily news and career tips to monthly insights on AI, sustainability, software, and more—pick what matters and get it in your inbox. Explore The Most Powerful Tech Event in the World with Interesting Engineering. Stick with us as we share the highlights of CES week! Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies. We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. Follow Us On Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. The research explores the use of audio-driven movement for humanoid robots. Chengxu Zhou, an associate professor in UCL Computer Science, has bagged an NVIDIA Academic Grant to support the latest endeavors to boost humanoid robots by focusing on real-time, audio-driven whole-body motion. The Robotics and AI professor will be supported with resources for both training and deployment through this grant. The resources include two NVIDIA RTX PRO 6000 GPUs and two Jetson AGX Orin devices. The addition of this result will enforce faster iteration and shorter training cycles, thereby reducing the time required between simulation and real-robot testing. NVIDIA awarded the grant as part of the Academic Grant Program, which provides compute resources to researchers working across different subject areas. Chengxu Zhou and his team are working on the Beat-to-Body project, which explores the possibilities of humanoids responding to sound with expressive, physically lausible, and safe whole-body movement. Instead of working with pre-scripted code, the system allows the humanoid robot to adapt to cues such as tempo, accents, and fluctuations in loudness, and to provide reactions. For instance, a consistent rhythm from clapping or music can guide a robot’s walking pace and body movements. At the same time, changes in the sound’s tone or intensity can create different styles of motion, ranging from quick, jerky steps to smooth, flowing movements. The goal is to enable the robot to react dynamically and adapt its behavior in real time as the audio evolves. The audio-first, on-robot execution plays a key role in this process. Researchers train at scale in simulation using GPU compute. In the meantime, the Jetson hardware enables low-latency inference directly on the robot, reducing reliance on offboard processing and enabling responsive reactions to sound cues. Zhou’s Beat-to-Body project aligns with a growing body of research exploring sound as a control signal for humanoid robots. A 2025 study demonstrated how robots can generate expressive locomotion and gestures directly from music and speech, without relying on predefined motion templates. Earlier work has shown that humanoids can synchronize dance movements to musical rhythm and emotion. Complementary research on audio-visual tracking has also enabled robots to localize and respond to sound sources, underscoring audio as an emerging, intuitive interface for human–robot interaction. Within the field of robotics, the project advances work on expressive, full-body movement and more intuitive human–robot interaction. In the near term, the technology could be used in interactive installations and performance-based settings. In the longer term, it may enable basic coordination among multiple robots using common audio signals. The work is being conducted at UCL’s Humanoid Robotics Lab, which specializes in machine learning, control systems, and interaction for humanoid robots. The next phase of the project will focus on expanding simulation-based training and testing early closed-loop, sound-responsive behaviors on a real humanoid robot. Atharva is a full-time content writer with a post-graduate degree in media & amp; entertainment and a graduate degree in electronics & telecommunications. He has written in the sports and technology domains respectively. In his leisure time, Atharva loves learning about digital marketing and watching soccer matches. His main goal behind joining Interesting Engineering is to learn more about how the recent technological advancements are helping human beings on both societal and individual levels in their daily lives. Premium Follow

Add the following snippet to your HTML:<iframe frameborder='0' height='385' scrolling='no' src='https://www.hackster.io/HiwonderRobot/building-a-multimodal-ai-brain-for-the-jetarm-robot-8accdc/embed' width='350'></iframe> See how the JetArm platform fuses vision, speech, and LLMs for autonomous decision-making—a practical guide to embodied AI. Read up about this project on See how the JetArm platform fuses vision, speech, and LLMs for autonomous decision-making—a practical guide to embodied AI. Robotic arms are evolving beyond simple pre-programmed machines into systems capable of interpreting and responding to their environment. This project explores the implementation of a multimodal AI decision-making pipeline on a robotic manipulator, using a platform like JetArm as a practical testbed. We'll break down how integrating vision, speech, and large language models (LLMs) can enable more autonomous and interpretable task execution. The core of this advanced functionality is a system that fuses multiple sensory inputs, acting as the robot's "eyes, " "ears, " and "reasoning center." The true innovation lies not in the individual components, but in their orchestration. Let's trace the decision pipeline through a concrete example: "Keep the item that is the color of the sky, and remove the others." 1. Intent Understanding & Semantic Grounding The spoken command is converted to text via ASR and sent to the LLM. The model must interpret the natural language. It deduces that "the color of the sky" refers to "blue." The core intent is extracted: classify objects by color, retain blue ones, remove others. This step bridges human language and machine-actionable goals. 2. Task Planning & Scene Analysis The system now correlates this intent with the visual scene. The vision model provides a list of detected objects with their properties (color, location). The planner: 3. Motion Execution & Closed-Loop Control The high-level plan is converted into low-level actions. For each target object: This three-stage pipeline (Understand → Plan → Execute) allows the system to handle variations in object placement, phrasing of commands, and scene layout without being explicitly reprogrammed. For developers and students, implementing such a pipeline on a platform like JetArm offers profound learning opportunities: This project moves beyond theoretical AI and into embodied AI, where intelligence is evaluated by its ability to generate physical action in an unstructured environment. This multimodal architecture demonstrates a significant step toward more adaptable and intuitive robots. The JetArm serves as an ideal platform for prototyping these concepts due to its integrated sensors and ROS-based software framework. Hackster.io, an Avnet Community © 2026

AbstractNvidia’s Isaac GR00T N1.6 is an advanced Vision-Language-Action (VLA) foundation model built specifically to empower humanoid robots with the ability to see, understand language, reason, and act in complex environments. Announced at CES 2026, GR00T N1.6 advances the company’s push toward “physical AI” — where robots don’t just perceive the world but learn generalized skills and perform coordinated actions. It extends Nvidia’s open robotic ecosystem, integrating perception, reasoning, and full-body control for real-world tasks. (NVIDIA Newsroom)What Is Isaac GR00T N1.6?Isaac GR00T N1.6 is an open vision-language-action (VLA) model designed as a foundation model for humanoid robots. It merges multimodal inputs — such as visual data and natural language instructions — and outputs continuous control actions that direct robot motion and task execution. As a VLA model, GR00T N1.6 enables robots to understand their surroundings and perform tasks with context, adaptability, and generalization across environments. (GitHub)At its core, a VLA model like GR00T combines:Vision: Interpreting camera images and surroundings.Language: Understanding instructions given in natural language.Action: Producing robot motion commands that accomplish tasks. (Wikipedia)Why GR00T Matters for Humanoid RoboticsTraditional robot systems rely heavily on pre-programming specific actions for specific tasks. GR00T N1.6 breaks from this by enabling flexible, generalized behavior:Cross-embodiment adaptability: The model can be fine-tuned for different robot bodies and configurations. (GitHub)Multimodal understanding: It integrates vision and language, allowing intuitive human interactions with robots. (Hugging Face)General task learning: Trained on diverse robot data (real and synthetic), enabling performance across varied tasks without bespoke coding. (Hugging Face)Full-body control: Supports coordinated movement and manipulation tailored to humanoid kinematics. (TechCrunch)These capabilities represent a leap toward generalist robots that can reason, plan, and act autonomously in dynamic human environments.How GR00T N1.6 WorksArchitecture OverviewGR00T N1.6 uses a hybrid architecture that combines:A vision-language foundation model (vision transformer plus language understanding)An action generation module (typically a diffusion or flow-matching transformer)Integration mechanisms that transform multimodal observations and instructions into executable control signalsThis structure enables seamless mapping from perception and instructions to robot actions. (Hugging Face)Training and DataThe model is trained on massive datasets comprising:Humanoid and bimanual robot trajectoriesSemi-humanoid dataSynthetic data generated through simulation toolsThis multimodal training enhances its ability to generalize to novel tasks. (Hugging Face)Typical WorkflowA typical usage scenario involves these steps:Collect robot demonstration data (video + state + action).Convert data to a compatible schema (e.g., LeRobot dataset format).Validate zero-shot performance using pretrained policies.Fine-tune the model for specific embodiments or tasks.Deploy the GR00T policy to run on robotic hardware. (GitHub)Use CasesThanks to its generalist capabilities, GR00T N1.6 is already being adopted across industries:Humanoid research platforms for testing advanced behavior and coordination. (NVIDIA Newsroom)Simulation-to-real workflows to train and validate robotics policies. (GitHub)Enterprise robotics solutions requiring natural language interaction and physical manipulation. (NVIDIA Blog)Education and research for robotics experimentation and custom task learning. (Hugging Face)Partners like Franka Robotics, NEURA Robotics, and other developers are leveraging GR00T-enabled workflows to simulate, train, and refine humanoid behaviors before production deployment. (NVIDIA Newsroom)Challenges and ConsiderationsWhile GR00T N1.6 represents a major advance, real-world deployment still necessitates:Robust hardware integration: Bridging simulation and real robot mechanics.Safety and ethics: Ensuring safe actions in human environments.Data collection quality: High-quality multimodal data remains essential for fine-tuning.These considerations require careful engineering and domain expertise for practical applications.The Future of Humanoid Physical AINvidia is positioning its robotics stack — including Cosmos Reason, simulation tools, and Jetson robotics hardware — as a comprehensive ecosystem for “physical AI,” where robots perceive, reason, and interact autonomously. The company aims to make robotics development more accessible and standardized, similar to how Android catalyzed mobile app ecosystems. (Unite.AI)The broader trend in robotics — integrating large multimodal foundation models with control systems — signals a shift from task-specific programming to generalized autonomous behavior. Models like GR00T N1.6 are at the forefront of this transformation.ConclusionIsaac GR00T N1.6 is Nvidia’s next-generation Vision-Language-Action model built specifically for humanoid robots. By integrating multimodal perception, language understanding, reasoning, and motion control, it empowers robots with generalized capabilities that go beyond scripted behaviors. As part of Nvidia’s expanding physical AI platform, GR00T N1.6 represents a meaningful step toward intelligent, adaptable humanoid robots poised to work alongside humans in unstructured environments — from research labs to real-world applications. Nvidia’s Isaac GR00T N1.6 is an advanced Vision-Language-Action (VLA) foundation model built specifically to empower humanoid robots with the ability to see, understand language, reason, and act in complex environments. Announced at CES 2026, GR00T N1.6 advances the company’s push toward “physical AI” — where robots don’t just perceive the world but learn generalized skills and perform coordinated actions. It extends Nvidia’s open robotic ecosystem, integrating perception, reasoning, and full-body control for real-world tasks. (NVIDIA Newsroom) Isaac GR00T N1.6 is an open vision-language-action (VLA) model designed as a foundation model for humanoid robots. It merges multimodal inputs — such as visual data and natural language instructions — and outputs continuous control actions that direct robot motion and task execution. As a VLA model, GR00T N1.6 enables robots to understand their surroundings and perform tasks with context, adaptability, and generalization across environments. (GitHub) At its core, a VLA model like GR00T combines: Vision: Interpreting camera images and surroundings. Language: Understanding instructions given in natural language. Action: Producing robot motion commands that accomplish tasks. (Wikipedia) Traditional robot systems rely heavily on pre-programming specific actions for specific tasks. GR00T N1.6 breaks from this by enabling flexible, generalized behavior: Cross-embodiment adaptability: The model can be fine-tuned for different robot bodies and configurations. (GitHub) Multimodal understanding: It integrates vision and language, allowing intuitive human interactions with robots. (Hugging Face) General task learning: Trained on diverse robot data (real and synthetic), enabling performance across varied tasks without bespoke coding. (Hugging Face) Full-body control: Supports coordinated movement and manipulation tailored to humanoid kinematics. (TechCrunch) These capabilities represent a leap toward generalist robots that can reason, plan, and act autonomously in dynamic human environments. GR00T N1.6 uses a hybrid architecture that combines: A vision-language foundation model (vision transformer plus language understanding) An action generation module (typically a diffusion or flow-matching transformer) Integration mechanisms that transform multimodal observations and instructions into executable control signals This structure enables seamless mapping from perception and instructions to robot actions. (Hugging Face) The model is trained on massive datasets comprising: Humanoid and bimanual robot trajectories Semi-humanoid data Synthetic data generated through simulation toolsThis multimodal training enhances its ability to generalize to novel tasks. (Hugging Face) A typical usage scenario involves these steps: Collect robot demonstration data (video + state + action). Convert data to a compatible schema (e.g., LeRobot dataset format). Validate zero-shot performance using pretrained policies. Fine-tune the model for specific embodiments or tasks. Deploy the GR00T policy to run on robotic hardware. (GitHub) Thanks to its generalist capabilities, GR00T N1.6 is already being adopted across industries: Humanoid research platforms for testing advanced behavior and coordination. (NVIDIA Newsroom) Simulation-to-real workflows to train and validate robotics policies. (GitHub) Enterprise robotics solutions requiring natural language interaction and physical manipulation. (NVIDIA Blog) Education and research for robotics experimentation and custom task learning. (Hugging Face) Partners like Franka Robotics, NEURA Robotics, and other developers are leveraging GR00T-enabled workflows to simulate, train, and refine humanoid behaviors before production deployment. (NVIDIA Newsroom) While GR00T N1.6 represents a major advance, real-world deployment still necessitates: Robust hardware integration: Bridging simulation and real robot mechanics. Safety and ethics: Ensuring safe actions in human environments. Data collection quality: High-quality multimodal data remains essential for fine-tuning.These considerations require careful engineering and domain expertise for practical applications. Nvidia is positioning its robotics stack — including Cosmos Reason, simulation tools, and Jetson robotics hardware — as a comprehensive ecosystem for “physical AI,” where robots perceive, reason, and interact autonomously. The company aims to make robotics development more accessible and standardized, similar to how Android catalyzed mobile app ecosystems. (Unite.AI) The broader trend in robotics — integrating large multimodal foundation models with control systems — signals a shift from task-specific programming to generalized autonomous behavior. Models like GR00T N1.6 are at the forefront of this transformation. Isaac GR00T N1.6 is Nvidia’s next-generation Vision-Language-Action model built specifically for humanoid robots. By integrating multimodal perception, language understanding, reasoning, and motion control, it empowers robots with generalized capabilities that go beyond scripted behaviors. As part of Nvidia’s expanding physical AI platform, GR00T N1.6 represents a meaningful step toward intelligent, adaptable humanoid robots poised to work alongside humans in unstructured environments — from research labs to real-world applications. ©2026 C# Corner. All contents are copyright of their authors.

From daily news and career tips to monthly insights on AI, sustainability, software, and more—pick what matters and get it in your inbox. Explore The Most Powerful Tech Event in the World with Interesting Engineering. Stick with us as we share the highlights of CES week! Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies. We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. Follow Us On Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. AI-driven scheduling enables agile output, producing excavators every six minutes and supporting fast, multi-model, small-batch manufacturing. A Chinese construction and agricultural machinery firm, Zoomlion, is accelerating its digital transformation by integrating AI across construction machinery, manufacturing, management, and robotics. Since 2024, Zoomlion has ventured into embodied-intelligence humanoid robotics, building on its fully integrated in-house development capabilities. Now, the company’s full-chain AI system aims to reshape operations into a fully intelligent enterprise, while its humanoid robots are positioned as a “third growth curve” supported by proprietary hardware-software integration. According to the firm, Zoomlion Smart City in China’s Hunan province operates smart factories under an end-to-end AI-driven manufacturing system. Zoomlion Heavy Industry has implemented a full-chain AI system spanning construction machinery, intelligent manufacturing, management, and humanoid robotics. At its Zoomlion Smart City, the company operates 12 smart factories with over 300 production lines, including 20 fully automated “lights-out” lines, forming an end-to-end intelligent manufacturing network. In the manufacturing segment, processes such as cutting, welding, machining, painting, and assembly are connected to an industrial internet platform, allowing coordinated management of more than 100,000 material types and production of over 400 distinct products. AI-driven scheduling and optimization support a high level of operational efficiency. The park produces an excavator every six minutes, a scissor lift every 7.5 minutes, a concrete pump truck every 27 minutes, and a truck crane every 18 minutes, demonstrating a move toward agile, multi-variety, small-batch production. Beyond manufacturing, Zoomlion applies AI to research and development, production planning, sales, service, and supply chain management. A voice-based AI diagnostic system provides technical support with over 95 percent accuracy, enabling remote fault detection, faster troubleshooting, and round-the-clock service. According to the firm, the integration of AI across operations reflects a broader industry trend toward data-driven manufacturing and digital management, emphasizing efficiency, flexibility, and responsiveness to demand rather than traditional large-scale, single-product production models. Zoomlion showcased its latest advancements in industrial and humanoid robotics at the 2025 World Robot Conference in Beijing, highlighting its dual role as a robotics developer and integrator. The company emphasized that robots are designed as collaborative partners rather than replacements, aiming to improve manufacturing quality, efficiency, and adaptability. Zoomlion’s robotics program dates back to 2006 with programmable industrial robots for large-scale, single-product production. Since 2019, the company has deployed adaptive robots equipped with vision and force sensing, integrated through an industrial internet platform to enable agile, multi-model production. At the Zoomlion Smart Industry City in Changsha, over 2,000 adaptive robots operate across 300 intelligent lines, producing cranes, excavators, aerial work platforms, and concrete machinery, with rapid model switching for high-mix, low-volume output. In 2024, Zoomlion expanded into humanoid robots, developing two-wheeled models and one bipedal unit, now piloted in machining, logistics, assembly, and quality inspection. These robots feature multimodal perception, intent recognition, integrated vision, force and tactile sensing, and dual-arm collaborative motion planning with safety awareness. The company supports deployment with a training facility of over 100 workstations and an AI-native cloud platform for large-scale data collection and model training. Connected through Zoomlion’s industrial internet platform, which spans 1.7 million units of equipment worldwide, these systems enable real-time coordination and adaptive production. Looking forward, Zoomlion plans to advance robot clusters that can self-perceive, adapt, and make decisions within smart factory environments. Jijo is an automotive and business journalist based in India. Armed with a BA in History (Honors) from St. Stephen's College, Delhi University, and a PG diploma in Journalism from the Indian Institute of Mass Communication, Delhi, he has worked for news agencies, national newspapers, and automotive magazines. In his spare time, he likes to go off-roading, engage in political discourse, travel, and teach languages. Premium Follow

Please ensure that JavaScript is enabled in your browser to view this page. LEGO building blocks may be the greatest toy ever conceived by humanity, but they have been missing something for the past few years. Since the discontinuation of the Mindstorms robotics kits, people who want to program their LEGO creations so they can do more than sit on a shelf to be admired have been out of luck. The recently unveiled SMART Brick will help to fill this gap, but this new offering does not have the capabilities most hobbyists are looking for. What we want is total hackability, and some substantial computing power to run AI or other powerful algorithms on LEGO bricks. A Kickstarter campaign created by STEMpedia seeks to give us exactly that. They have created Quarky Intellio, which is a LEGO- and Arduino-compatible kit for adding AI, AR, and more to your LEGO creations. It is a special block with LEGO mounts that contains an ESP32-S3 microcontroller, sensors, and interfaces for integrating with motors and other electronic components. The Quarky Intellio module integrates an ESP32-S3, 5-megapixel autofocus camera, microphone, speaker, SD card slot, TFT display interface, GPIOs, servo ports, and a 1,000 mAh rechargeable battery into a single, compact unit. This makes it possible to build self-contained, mobile projects without a tangle of external boards or power supplies. Because it is based on the ESP32-S3, Intellio is capable of running edge AI workloads locally, avoiding the need for constant cloud connectivity. On the software side, Intellio is designed to grow with the user. Beginners can start with block-based programming in PictoBlox, while more advanced users can switch to Python or C++ through the Arduino IDE, ESP-IDF, or MicroPython. Machine learning models can be trained using PictoBlox’s ML environment or deployed with TensorFlow Lite. This flexibility enables projects such as autonomous POV cars, AR-enabled robots that recognize AprilTags, object-tracking soccer bots, and interactive desk companions that combine face recognition with speech input. The module can interface with popular platforms like Arduino, Raspberry Pi, and ESP32-based boards, as well as a wide range of servos, sensors, and displays to enable more complex functions. Two main kits are offered: the Discovery Kit, which introduces AI and AR concepts, and the more advanced Rover Kit, which adds motors and hardware for building fully autonomous robots. This campaign closes in just a few days, so if you want to get in on the rewards that start at $49, don’t wait too much longer. Full details are available at Kickstarter. Hackster.io, an Avnet Community © 2026

From daily news and career tips to monthly insights on AI, sustainability, software, and more—pick what matters and get it in your inbox. Explore The Most Powerful Tech Event in the World with Interesting Engineering. Stick with us as we share the highlights of CES week! Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies. We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. Follow Us On Access expert insights, exclusive content, and a deeper dive into engineering and innovation. Engineering-inspired textiles, mugs, hats, and thoughtful gifts We connect top engineering talent with the world's most innovative companies We empower professionals with advanced engineering and tech education to grow careers. We recognize outstanding achievements in engineering, innovation, and technology. All Rights Reserved, IE Media, Inc. A rare multilingual AI humanoid robot conversation powered entirely on device. At CES 2026, humanoid robotics company Realbotix publicly demonstrated what it described as one of the first fully autonomous, unscripted conversations between two physical humanoid robots powered entirely by embedded AI. The interaction, which took place live on the show floor, involved two humanoid robots, Aria and David, engaging in a continuous, real-time dialogue for more than two hours without human intervention, scripting, or teleoperation. According to the company, both robots ran Realbotix’s proprietary AI software locally on-device, rather than relying on cloud-based processing. The demonstration was positioned as an example of “physical AI,” with two embodied systems perceiving, responding, and adapting to each other in real time rather than following pre-programmed dialogue trees. “Realbotix has specialized in robots for human interaction. In this case, we demonstrated that our robots can interact with each other,” said Andrew Kiguel, CEO of Realbotix. “This is a true demonstration of physical AI in action, as the interaction was completely unscripted and lasted over two hours.” During the demonstration, Aria and David conducted their conversation across multiple languages, including English, Spanish, French, and German. Realbotix said this highlighted both the multilingual capabilities of its language models and the flexibility of its embodied AI platform. The exchange unfolded organically, with the robots taking their time to respond to each other’s statements rather than following a fixed conversational structure. In one moment, a robot remarked, “No coffee jitters and no awkward pauses and only silicon charisma,” a line that seems to have emerged from the system’s conversational flow rather than a scripted prompt. However, observers noted that the live interaction featured noticeable pauses, speech inconsistencies, and uneven pacing. Visually and expressively, the robots’ delivery remained limited and certainly not comparable to high-profile humanoid demonstrations such as Ameca, which has disturbingly real facial expressions, facial animations, and fluidity. Even compared to everyday AI voice assistants like GPT-4o, Realbotix’s humanoids appeared more mechanical, with restrained expressions and speech. Online viewers also characterized the robots as resembling “rubber mannequins with speakers.” Alongside the humanoid-to-humanoid exchange, Realbotix presented a separate demonstration focused on human-robot interaction. A third humanoid robot demonstrated the company’s patented vision system, embedded in the robot’s eyes. The company says that during this demo, the robot interacted verbally with attendees, identifying individuals, tracking them visually, and interpreting voice and facial cues to infer emotional states. Realbotix said the vision system enabled the robot to follow people naturally and respond in a conversational manner, highlighting progress in real-time visual perception and embodied social interaction. While the conversational quality was not seamless, the demonstration’s significance lay elsewhere. Most humanoid robot showcases rely on tightly controlled environments, teleoperation, or pre-scripted dialogues designed to minimize errors. In contrast, Realbotix allowed its systems to operate openly, exposing limitations such as pauses, timing mismatches, and uneven speech delivery. Rather than presenting a choreographed performance, Realbotix showed how two autonomous humanoid systems currently behave when left to interact freely, in public, and for an extended duration. Realbotix designs and manufactures AI-powered humanoid robots intended for entertainment, customer service, and companionship. The company says its robots are manufactured in the United States and use patented technologies to enable lifelike expressions, motion, vision, and social engagement. By staging the demonstration at CES 2026, Realbotix placed its technology in front of industry leaders, investors, and media for an unfiltered look at the current state of their embodied, on-device conversational AI. Kaif Shaikh is a journalist and writer passionate about turning complex information into clear, impactful stories. His writing covers technology, sustainability, geopolitics, and occasionally fiction. A graduate in Journalism and Mass Communication, his work has appeared in the Times of India and beyond. After a near-fatal experience, Kaif began seeing both stories and silences differently. Outside work, he juggles far too many projects and passions, but always makes time to read, reflect, and hold onto the thread of wonder. Premium Follow

Beijing, Jan. 11 (SANA) Chinese firms dominated the global market for human-like robots in 2025, both in shipments and market share, according to a new report by research company Omdia. The report, “Radar of the Embodied Robotics Market for General Purposes,” cited by Xinhua, said the industry entered a phase of rapid growth last year, with total annual shipments expected to reach around 13,000 units. The market leader was Chinese technology innovator Agibot, which shipped 5,168 units in 2025, accounting for 39% of the global market, followed by Unitary Technology. The study highlighted the growing integration of generative artificial intelligence with robotics. This development is driving the creation of “general embodied intelligence,” robots that can learn and adapt to a variety of environments. Looking ahead, the human-like robot market is expected to expand significantly over the next decade, with global shipments projected to reach about 2.6 million units by 2035. Last month, China also launched the world’s first global app store dedicated to human-like robots, signaling further growth and innovation in the sector. M.S Editors Choice Syrian Arab News Agency – SANA The official national news agency of Syria, established on June 24, 1965. It is affiliated with the Ministry of Information and headquartered in Damascus. Sign in to your account Remember me

Robots are getting better at seeing, hearing, and moving, but touch has always been the missing piece. At CES 2026, Ensuring Technology showcased a new kind of artificial skin that could finally give robots something close to human-like sensitivity, helping them feel the world instead of just bumping into it. The company’s latest tactile sensing tech is designed to help robots understand pressure, texture, and contact in ways that go beyond simple touch sensors. At the center of the announcement are two products called Tacta and HexSkin, both aimed at solving a long-standing problem in robotics. Humans rely heavily on touch to grasp objects, apply the right amount of force, and adapt instantly when something slips. Robots, on the other hand, usually operate with limited feedback. Ensuring Technology’s goal is to close that gap by recreating how human skin senses and processes touch. Tacta is a multi-dimensional tactile sensor designed for robotic hands and fingers. Each square centimeter packs 361 sensing elements, all sampling data at 1000Hz, which the company says delivers sensitivity on par with human touch. Despite that density, the sensor is just 4.5mm thick and combines sensing, data processing, and edge computing into a single module. At CES, Ensuring demonstrated a fully covered robotic hand using Tacta, with 1,956 sensing elements spread across fingers and palm, effectively creating a complete network of tactile awareness. HexSkin takes the idea further by scaling touch across larger surfaces. Built with a hexagonal, tile-like design, HexSkin can wrap around complex curved shapes, making it suitable for humanoid robots. CES 2026 has been packed with robots that show just how fast the field is moving, and why better touch matters. We’ve seen LG’s CLOiD home robot pitched as a household helper for chores like laundry and breakfast, alongside humanoid robots that can play tennis with impressive coordination and Boston Dynamics’ Atlas, which displayed advanced balance and movement this time. While these machines already see and move remarkably well, most still rely heavily on vision and rigid sensors. Adding a human-like touch through artificial skin could be what finally makes robots feel a little more human. Anthropic just stepped into the healthcare AI space with the launch of Claude for Healthcare, a new suite of tools designed for providers, payers, and patients. Following in the footsteps of OpenAI's ChatGPT Health, Claude for Healthcare aims to bring AI safely into medical contexts, helping users access and understand their health information more effectively. As part of this push, Anthropic is introducing new integrations that let users connect their health data to Claude. In the US, subscribers on the Claude Pro and Max plans can give the AI assistant secure access to lab results and health records, and unlock features that make that data actionable. A team of researchers in China has just pulled the curtain back on a new sodium-sulfur battery design that could fundamentally change the math on energy storage. By leaning into the very chemistry that has historically made sulfur a headache for engineers, they have managed to build a cell that is incredibly cheap to make but still packs a massive energy punch. The design, which is currently being tested in the lab, uses dirt-cheap ingredients: sulfur, sodium, aluminum, and a chlorine-based electrolyte. In early trials, the battery hit energy densities over 2,000 watt-hours per kilogram - a figure that blows today’s sodium-ion batteries out of the water and even gives top-tier lithium cells a run for their money. One of the most interesting updates in the solar energy space comes from the Hebrew University of Jerusalem (via EES Solar), where researchers have developed 3D-printed, semi-transparent, flexible perovskite solar cells with adjustable color and transparency. Traditionally, solar panels are either blue, dark gray, or black, depending on the type of panel and the materials used to make them. While that uniform look works well on rooftops where nobody really sees them, visible solar installations often fail to complement architecture or design. This issue affects urban residents not just in the United States, but across the globe. Upgrade your lifestyleDigital Trends helps readers keep tabs on the fast-paced world of tech with all the latest news, fun product reviews, insightful editorials, and one-of-a-kind sneak peeks.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript. Advertisement Nature Communications volume 16, Article number: 6073 (2025) Cite this article 12k Accesses 1 Citations 64 Altmetric Metrics details Insect–computer hybrid robots offer strong potential for navigating complex terrains. This study identified the intersegmental membrane between the pronotum and mesothorax of the Madagascar hissing cockroach as an effective site for electrical stimulation to control direction and speed. A pair of bipolar electrodes was custom-designed, and an automatic assembly system was developed, integrating a robotic arm, vision-based site detection, and an insect fixation structure. The system achieved assembly in 68 s. Hybrid robots exhibited robust steering (over 70°) and deceleration (68.2% speed reduction) with performance comparable to manually assembled counterparts. Controlled navigation along an S-shaped path confirmed accurate directional control. Furthermore, a multi-agent system of four hybrid robots covered 80.25% of an obstructed terrain in 10 minutes and 31 seconds. This work demonstrates a scalable strategy for automating the fabrication of insect–computer hybrid robots, enabling efficient and reproducible assembly process while maintaining effective locomotion control. Various insect-scale robots have been engineered, demonstrating exceptional maneuverability within complex and narrow terrains1,2,3,4,5. This capability has spurred advancements in mechanically structured and insect-computer hybrid robots (i.e., biobots or cyborg insects). While these robots have a similar size, they possess unique self-locomotion energy sources3,6,7 and adaptability to challenging terrains8. Therefore, such robots’ potential is increasingly explored as a robotic platform3,9,10,11,12,13 across various applications. To achieve locomotion control in insects, researchers have studied stimulation electrodes targeting their muscles, neuron systems, and sensory organs7,14,15,16,17. Both invasive2,3,15 and non-invasive electrodes7 have been manually implanted into the insects’ target body parts to enhance stimulation. However, the intricate and fragile anatomy of insects makes the manual surgery process time-consuming (~15 min per insect18) and difficult15. For instance, insects’ small and soft antennae and cerci necessitate the use of specialized microinstruments and microscopes to manipulate their tissues accurately. Even minimal force can cause unintended shape changing, making procedures highly demanding (Supplementary Fig. S1). Furthermore, the success of the surgical procedure significantly reliant on the operator’s expertise2,3,15, which impacts the risk of accidental injury to the insects. Due to variations in the operator’s surgical skills, even with the same implantation method, the insect–computer hybrid robots operated by different operators may behave differently. For consistent production of insect-computer hybrid robots, transitioning from manual to automatic assembly processes is imperative, particularly for applications that demand large scale deployment, such as post-disaster search and rescue or factory inspections, where multiple robotic agents significantly enhance efficiency than a single unit19,20. Therefore, developing automatic assembly methods for insect-computer hybrid robots is crucial for mass production. Madagascar hissing cockroach has been used in various applications as a powerful platform3,8,13,21. Consequently, our study focuses on optimizing its mass production. Although these insects generally share a similar body structure, individual size variations pose challenges to achieving uniform localization of the implantation site in contrast to some assembly tasks for mechanical parts of certain shapes22. Consequently, to ensure precise localization of the implantation site, advanced deep learning techniques are employed. To control the locomotion direction of Madagascar hissing cockroach, stimulation protocols have been developed on the insects’ antennae and abdomen3,15,23. The antennae detect and navigate around the obstacles24, and the tactile stimuli25 significantly influence cockroach movement. The electrical stimulation of the antennae effectively induces directional turning among the insects7,15. However, the antennae are soft, fragile, and tiny15 (with a 0.6–0.7 mm diameter, Fig. 1B), posing challenges in securely attaching and implanting electrodes. Moreover, antennae used in hybrid robots to enable autonomous obstacles navigation7,26 destroy the insects’ innate ability to maneuver around the obstacles7,27. Therefore, this study excludes antennae as the target site for stimulation. A An anesthetized cockroach secured for backpack assembly using the robotic arm. The automatic assembly included an insect fixation structure driven by a slide motor, a robotic arm equipped with a gripper for grasping and assembling the backpack, and a depth camera for precise localization of the insect’s body position during assembly. B Components of insect-computer hybrid robot. The white mounting structure was attached with a microcontroller and bipolar electrodes. The grasping section of the backpack was held by a robotic arm gripper, while the alignment hole placed the backpack. Mounting branches were used to hook the insect’s metathorax. Bipolar electrodes of the backpack were implanted into the intersegmental membrane between the insect’s pronotum and mesothorax. C Insect-computer hybrid robot. The backpack-assembled insect was controlled to execute turning maneuvers and decelerating. D A microcontroller with Sub-1GHz communication used to between the insect-computer hybrid robot and workstation. Stimulation signals were transmitted from the stimulation channels to the bipolar electrodes for insects’ locomotion control. After the assembly finished and the hybrid robot was needed for the locomotion control, a LiPo battery was plugged inside the power socket. Double-sided tapes were used to stick the battery to the backpack. Stimulating the abdomen’s sides influences the locomotion direction of the insects3. However, cockroaches’ abdominal cuticles are short and thin—with the third abdominal cuticle measuring 3.8–5.0 mm in length and 0.2–0.3 mm in thickness—making automatic electrode implantation difficult. Consequently, an alternative stimulation site should be explored. Building upon studies on similar terrestrial platforms such as Zophobas morio, which alters its locomotion direction when stimulated on the pronotum and elytra2, we hypothesized that pronotal stimulation would similarly alter the cockroach’s direction. Notably, the pronotum cuticle is both larger and thicker than the abdominal cuticles, measuring ~11.6–13.4 mm in length, 0.5–0.6 mm in thickness, which eases attachment and detection. Additionally, akin to the interspace between abdominal segments7, an intersegmental membrane between the pronotum and the mesothorax is the key target site for stimulation, enabling directional control of the insect (Fig. 1C). This study proposes a stimulation protocol designed to control an insect-computer hybrid robot. To analyze the insects’ responses to electrical stimulation, we recorded their neural activities, foreleg movements, and locomotion patterns during simulation. A backpack was developed, integrating microcontroller, stimulation electrodes, and a mounting device (Fig. 1B, C). The hybrid robot was controlled wirelessly for steering and stopping. Next, an automatic assembly was developed, incorporating a slide motor, a fixation structure for the insect, an intel RealSense D435 camera, a Robotiq Hand-e gripper and a Universal Robot UR3e (Fig. 1A). This system relied on the visual detection of the target body position. During the assembly process, the insect was fixed in a structure mounted on a motor-driven slider. After assembling the backpack into the insect, the insect fixation structure was released, completing the assembly of the hybrid robot. Five automatically assembled hybrid robots were studied for locomotion control, including steering, and deceleration, with their performance compared to that of the manually assembled ones. Four hybrid robots traversed an outdoor uneven terrain using a UWB localization system. Due to the challenges of securing the antennae and abdomen and implanting electrodes with the robotic arm, this study focuses solely on the pronotum. To facilitate turning locomotion in cockroaches and minimize the number of implantation sites, a pair of bipolar electrodes was designed and implanted on the left and right sides of the intersegmental membrane between the pronotum and mesothorax (Figs. 1C and 2A). Each bipolar electrode comprised a copper pattern for electrical signal transmission and a microneedle structure to rapidly puncture the intersegmental membrane, and a hook mechanism to prevent detachment after implantation. A Custom-designed bipolar electrode utilized a microneedle structure integrated with a hook design, facilitating rapid membrane penetration and secure self-locking within the punctured membrane. The electrode comprised normal resin and patterned copper wires to transmit stimulation. B Bipolar electrodes fabricated using multi-material 3D printing technology followed by electroless plating. C Bipolar electrodes before and after copper plating. D Finite element modeling and analysis during the process of bipolar electrodes implantation into the intersegmental membrane between the cockroach’s pronotum and mesothorax. E Adhesion rating of the metal plating within the bipolar electrode, evaluated using the ASTM D3359-09 standard. Incorporating chemical etching strengthened the metal adhesion, preventing detachment during implantation and use and preserving electrical performance. F Impedance of conductive bipolar electrodes with conductivity (<70 Ω), significantly lower than non-invasive electrodes’ impedance7. G Plating thickness vs. time. The curve depicts a growth trend, representing a thicker coating layer with a longer plating duration. The complex structural features and the combination of plastic and metal components in bipolar electrodes necessitated specialized fabrication processes. Integrating multi-material 3D printing technology with an electroless plating process offered an effective approach for fabricating 3D electronic structures with spatial configuration and electrical signal carrying functions28,29,30 (Fig. 2B, Supplementary Movie S1). Initially, multi-material DLP3D printing was employed to fabricate the precursor structure of the bipolar electrodes, integrating a normal resin with an active precursor. The active precursor contains a catalytic agent that facilitates selective metal deposition during electroless plating, allowing metallization on both sides of the structure. Figure 2C presents the precursor and the final bipolar electrode with selectively deposited copper. To ensure effective implantation and electrical stimulation, the bipolar electrodes should possess high hardness and toughness. Therefore, ABS-like photosensitive resins31 was chosen as the material for electrode fabrication (in their normal resin form or as an active precursor). Finite element simulations (Fig. 2D, i–iv) showcased bipolar electrodes exhibiting a uniform stress distribution and controlled deformation during implantation, without experiencing damage or yielding. Additionally, the designed bipolar electrodes were securely implanted by monitoring the microtip’s stabbing stress at the membrane (Fig. 2D, v). To address the delamination or separation of metal plating during implantation, we calibrated the plating adhesion by the ASTM D3359-09 standard. We incorporated chemical etching into the selective electroless plating to enhance the adhesion and achieve a high 4B grade (Fig. 2E). Additionally, the implanted part comprising cockroach’s soft tissue (intersegmental membrane) and electrolyte solution, exerted minimal cutting force on the electrode, reducing impact (Fig. 2A). The impedance profile (Fig. 2F) of a plated layer on the bipolar electrode reveals that the impedance remained consistently below 70 Ω, considerably lower than the impedance of comparative non-invasive electrodes (exceeding 1000 Ω7). This reduced impedance facilitated stronger stimulation and a more pronounced insect reaction. The electrical conductivity and the thickness of the plating were 3.12 × 107 S/m and 2.5 µm, respectively. The copper-replacing-nickel plating method caused the electrodes to achieve selective copper metallization. After 5 min of immersion in the plating solution, a substantial increase in plating thickness was observed (Fig. 2G). A thicker plating layer enhanced conductivity, reduced impedance, and improved corrosion resistance; however, it increased parasitic capacitance. This study selected a plating time of 16 min for plating thickness and conductivity optimization. By controlling the electroless plating’s duration, the plated layer’s thickness on the bipolar electrodes was precisely modulated, enabling the fine-tuning of conductivity and other electrochemical properties. The width of the intersegmental membrane between pronotum and mesothorax was measured (1.4 ± 0.2 cm) across cockroaches with different sizes. To ensure that the bipolar electrodes implanted into the membrane, the distance between the two bipolar electrodes was set as 1.0 cm to satisfy minimum width of the membrane, i.e., 1.2 cm. To optimize stimulation voltage, neural activities in the insects’ neck region were recorded and analyzed (Fig. 3A). A progressive increase in the number of detected neural spikes was observed as the voltage increased from 0.5 V to 3.0 V, exhibiting heightened sensitivity to stronger electrical stimulation. A plateau at 3.0–3.5 V indicated that electrical stimulation above 3.0 V did not produce a strong neural response from the insects. However, increasing the stimulation voltage to 4.0 V yielded a 23.5% decline in the average number of spikes, denoting reduced neural activity due to potential damage to the insects’ neural system32,33,34. To prevent unnecessary damage and maintain effective stimulation, a 3.0 V voltage was considered optimal for the subsequent discussions. A Insect’s neural activity in response to electrical stimulation. i Neural recording configuration. Two probes were fixed to the nerve cord within the cockroach’s neck to record neural signals. ii Neural activity and corresponding electrical stimulation. Neural spikes induced by electrical stimulation (blue circles) were quantified to assess responses across various stimulation voltages. iii Spike counts at varying stimulation voltages (mean ± SD). For each voltage, nine trials of recording were conducted. B Insect leg reactions to electrical stimulation. i No stimulation: The insect’s forelegs remained extended. ii Right-turning stimulation: The insect’s left foreleg was stimulated and contracted. iii Left-turning stimulation: The insect’s right foreleg was stimulated and contracted. iv Deceleration stimulation: Both insect’s forelegs were stimulated and contracted. C Responses of insect locomotion to electrical stimulation. i Induced angular speed during turning stimulation (mean ± SD). ii Angular variations during turning stimulation (mean ± SD. iii Induced linear speed during deceleration stimulation (mean ± SD). Since the pronotum of the insect is connected to its forelegs (Fig. 3B) which guides its locomotion35, examining the forelegs’ status during stimulation is crucial. When one side of the pronotum was stimulated, the foreleg on that side contracted until the electrical stimulation was discontinued (Fig. 3B, Supplementary Movie S2), confirming that the stimulation directly influenced forelegs. Given that the forelegs guide the insects during walking35, this observation suggests their usage in controlling the insect’s orientation. Consequently, this study explored locomotion in detail. To simplify future automatic assembly processes, the stimulation bipolar electrodes, microcontroller, and mounting parts were integrated into a backpack. These backpacks were manually and then automatically affixed to insects to compare locomotion control. The manually assembled hybrid robots were tested for their locomotion control (N = 5 insects). The insects’ responses to the electrical stimulation applied to both sides of pronotum (Fig. 3C) confirmed that the implanted stimulation electrodes, positioned within the intersegmental membrane between pronotum and mesothorax, induced directional movement in the insects. Electrical stimulation turn the insects to the left and right with average angles of 68.0° and 82.6°, and maximum angular speeds of 275.8°/s and 298.2°/s, respectively. Additionally, the simultaneous contraction of the forelegs during stimulation (Fig. 3B) suggests that the implantation of the insect’s intersegmental membrane between the pronotum and mesothorax successfully stimulated the forelegs, with the insect steered with unilateral stimulation of a foreleg. Our stimulation protocol outperforms previous non-invasive electrode-based methods7. It enhances the maximum steering speed by over five times and increases the turning angle by over 76.6%. Additionally, this protocol requires only 40% of the stimulation time and 75% of the stimulation voltage. These enhancements indicate that our protocol triggered more intense turning responses with reduced time and energy consumption, signifying a more resource-efficient approach, improving the hybrid robot’s overall performance, operational efficiency, and cost-effectiveness. Our approach optimizes locomotion control for insect-computer hybrid robots, making it valuable for practical applications that demand quick and energy-efficient manoeuvrability. Furthermore, the cockroach decelerated when the electrical stimulation was outputted from the outer electrodes of the bipolar electrode pair (Fig. 2A), making the first successful implementation of the deceleration control in insect-computer hybrid robots. For the control of the insect-computer hybrid robot, both direction control and speed control are fundamental. Previously designed direction control, including left and right turning speed realized the full direction control of the insect-computer hybrid robot. However, for the speed control, only acceleration stimulation was realized. Hence, our discovery on the deceleration control could realize the speed control more fully. After 0.33 s of stimulation, the insects’ average walking speed declined from 6.2 cm/s to a minimum of 1.5 cm/s (Fig. 3C, iii), translating their body length/s to reduce from 112.9% to 27.6% with an average body length of 5.5 cm. Consequently, the insects experienced an 85.3% deceleration relative to their body length. The standard deviation of the minimum speed normalized with the mean speed during stimulation was 0.83. This small normalized standard deviation in the minimum decelerated speed highlights the consistency of the deceleration stimulation. The simultaneous contraction of both forelegs during stimulation (Fig. 3B, iv) further highlights the direct correlation of the deceleration trend in cockroaches with the stimulation of their forelegs. Our findings prove that the proposed stimulation protocol on the pronotum realized steering and deceleration control of the insect-computer hybrid robot (Supplementary Movie S3). The automatic assembly of hybrid robots comprises the following steps: 1) The pronotum and mesothorax of an anesthetized insect are secured and the intersegmental membrane is exposed. 2) A reference point is identified for electrode implantation. 3) The backpack is grasped using a robotic arm’s gripper. 4) The bipolar electrodes are implanted into the exposed membrane utilizing the robotic arm. 5) Force is applied to the backpack until its mounting branches latch onto the insect’s metathorax. 6) The backpack is disengaged from the gripper. 7) The robotic structure is retracted to release the insect (Fig. 4A). A Automated assembly. i, ii An anesthetized insect on the platform was secured with a 3D customized structure. iii Insect along with the detected reference point for bipolar electrode implantation (green dot). iv–vi The robotic arm grasping a backpack, implanting bipolar electrodes, and mounting the backpack. vii, viii The robotic arm releasing the backpack to allow the fixation structure to retract. B Insect fixation. i Configuration for insect fixation. ii Magnified view of the insect’s pronotum fixation. iii Details of the fixation with markings. Rods A and B pressed the anterior pronotum and the mesothorax, respectively, elevating the posterior pronotum. iv Pronotum lifting height h increased as the distance d of Rod A decreased. For each d value, lifting heights h of ten insects were recorded. The error bars denote the standard deviation. C Detection of the implantation reference point. i A square region around the pronotum of the insect cropped for segmentation inference. ii Pronotum of the insect identified using the TransUnet model. iii) Edge of pronotum was identified using the detected pronotum mask. iv Midpoint of the posterior pronotum edge detected as the implantation reference point. D Implantation of the bipolar electrode. i Measurement of the implantation pitch angle (α) while preventing collisions with the insect or the fixation structure. ii Recording of lower threshold αL and upper threshold αU when the backpack touched the 3D-customized structure and the insect, respectively. Five trials of the experiment were conducted. The error bars denote the standard deviation. To minimize the risk of potential collisions, the midpoint of these thresholds, α = 162.7°, was selected. Key considerations that must be addressed for successful assembly are as follows. First, the insect’s pronotum and mesothorax must be fixed to reveal their intersegmental membrane (Fig. 4A, ii). Second, the reference point on the pronotum must be identified for accurately implanting the bipolar electrodes by the robotic arm. Finally, the robotic arm must be maneuvered to assemble the backpack onto the insect at an optimal angle for secure assembly. The intersegmental membrane between the pronotum and mesothorax is concealed beneath the hard cuticle of the connected pronotum (Fig. 4B, ii). The posterior pronotum is elevated from the mesothorax to implant bipolar electrodes within this membrane (Fig. 4B, ii) using the developed Rod A and Rod B (Fig. 4B, iii). Rods A and B exerted pressure to the anterior pronotum and the mesothorax, respectively, exposing the membrane to enable electrodes implantation (Fig. 4B, iii). Initially, Rod A was positioned 4.0 mm above the platform, corresponding to a lowered distance, d, of 0 mm. The relationship between the lowered distance, d, and the lifting height, h, of the pronotum (Fig. 4B, iii) highlights the progressive exposure of the intersegmental membrane as the structure is lowered. The bipolar electrode’s thickness of 0.6 mm, required the lifting height, h, to consistently exceed this threshold to ensure sufficient space for implantation. On average, when d ≥ 1.5 mm, the height, h, reaches 1.9 mm which provides adequate exposure for successful implantation. For d ≥ 3.5 mm, h stabilizes around 1.9 mm, suggesting that further lowering the structure does not significantly increase membrane exposure (Student’s t test for d = 1.5 and 2.0 mm, and P = 0.31). As the lifting heights for both d = 1.5 mm and d = 2.0 mm surpass twice the thickness of the bipolar electrodes, d = 1.5 mm was chosen to prevent excessive pressure on the insect’s body while ensuring adequate membrane exposure for bipolar electrode insertion. This approach minimizes the potential injury to the insect, ensuring its physical integrity throughout implantation. To implant bipolar electrodes into the exposed intersegmental membrane (Section “Exposure of intersegmental membrane between pronotum and mesothorax”), a computer vision system was employed to determine the membrane’s location. Since bipolar electrodes on both sides of the backpack needed simultaneous implantation symmetric to the insect’s intersegmental membrane, a reference point pR was established at the middle point of the posterior pronotum edge. However, the pronotum varied in size and shape across each insect (Supplementary Fig. S5). Although the structure restrained the fixing of the insects’ pronotum and mesothorax, the pronotum’s position may still vary along the x- and y- axes (Fig. 4B, iii, Supplementary Fig. S5). Therefore, implementing a deep learning-based computer vision model is crucial for accurately identifying the pronotum and determining the location of the reference point, pR. An evaluation was conducted on several widely used segmentation models, including UNet36, Deeplabv337, TransUNet38, as well as recently developed segmentation models, such as Segment Anything39, Segment Anything240, LM-Net41, InceptionNeXt42, EMCAD43 and SHViT44, for their accuracy in pronotum segmentation. The models’ performance was measured by the mean intersection over union (mIoU) score, mean Dice similarity coefficient (mDSC), and mean squared error (MSE) of pR prediction (Table 1). The TransUNet model surpassed other models based on mIoU, mDSC and MSE metrics due to its hybrid encoder, which leverages the strengths of the Transformer architecture and preserves locality through its convolutional neural network (CNN). The other hybrid models, LM-Net, SHViT and EMCAD were designed to be lightweight models geared towards faster inference times, causing poor performance, apart from LM-Net, which had achieved comparable scores with the other models despite having significantly less parameters. The CNN-based models UNet, Deeplabv3 and InceptionNeXt managed to achieve better segmentation accuracies compared to the foundational ViT-based models SAM and SAM2, with Deeplabv3 even having similar performance to TransUNet in all three metrics as it was generally able to segment out most of the pronotum but did not precisely identify the lower border of the pronotum, causing slightly poorer scores on the three metrics. Subsequently, DSC loss and Boundary Difference Over Union (bDoU) loss45 were compared with BCE loss to achieve improved boundary results. The results of the loss function evaluation were shown in Table 2 in terms of mIoU score, mDSC score, and MSE of pR prediction. The DSC and BCE loss functions generally outperformed the bDoU loss function on the mIoU and mDSC metrics, as they evaluated the prediction the full pronotum while the bDoU loss focused primarily on the boundaries of the segmented objects. However, the bDoU loss function achieved the best MSE score among the three, highlighting its effectiveness in training the model to learn object boundaries. Ultimately, the chosen deep learning solution was the TransUNet model trained with the BCE loss function (Fig. 4C), as it had the best mIoU and mDSC scores and only slightly underperformed on the MSE metric compared to the model trained with bDoU loss. Data ablation study was also conducted to evaluate how the effect of different data augmentation methods on the training images impacts the segmentation performance. Performance based on mIoU, mDSC and MSE of pR prediction, is shown in Table 3. The TransUNet model had achieved significantly better scores across all three metrics when trained with augmented data, compared to when trained with only original data. The model trained on the asymmetrically scaled augmented data had shown to have slightly better performance in segmenting the boundaries compared to the model trained on the rotated augmented data, attributed to the robustness to variations in the shapes of pronotums. However, the latter had a better segmentation accuracy, as indicated by the larger mIoU and mDSC scores. To assemble the backpack onto the insect, the robotic arm scanned the insect with the camera and identified the reference point for bipolar electrode implantation (Section “Detection of pronotum using deep learning”). The Robotiq Hand-e gripper secured the grasp on the backpack and allowed stable electrode implantation with a gripping force of up to 185 N. Based on the combined weight of the Robotiq Hand-e (1.0 kg), RealSense D435 camera (75.0 g), and the backpack (2.3 g), the Universal Robots UR3e was selected as the robotic arm due to its 3.0 kg payload capacity and a pose repeatability of 0.03 mm at full payload46. Additionally, the UR3e’s 500 mm reach supports the RealSense D435’s minimum depth sensing distance of ~280.0 mm when operating at maximum resolution. To accurately identify the reference point for bipolar electrode implantation, the robotic arm was vertically oriented, with its gripper and camera positioned directly downward. Once the camera verified the reference point’s position, the robotic arm descended to the backpack’s position. The backpack was placed within the backpack holder (Fig. 1A), yielding consistent waypoints for the robotic arm to grasp the backpack. Upon grasping the backpack, the robotic arm transported it to the same pre-implant waypoint. Next, the robotic arm conducted bipolar electrodes implantation into the exposed intersegmental membrane. Due to the constrained manipulation space available for the robotic arm (6.5 × 3.5 × 2.5 cm3, between the insect and the structure), it was essential to determine an optimal implantation angle for the bipolar electrodes to avoid any potential collision. Given the symmetry of the insect’s pronotum and its alignment with the reference point (Fig. 4B, iii), only the pitch angle (α, rotation around y-axis) required adjustment while disregarding the roll and yaw angles. During the implantation, potential collisions can occur between the backpack and the 3D-designed structure, or between the backpack branches and the insect’s dorsal cuticles. Consequently, we identified pitch angles at which the backpack contacted the 3D structure (αL) and the insect’s dorsal cuticles (αU), utilizing five insects. Our measurements (Fig. 4D, ii) indicated αL to be 157.8° ± 1.5° and αU to be 167.5° ± 2.2° (mean ± standard deviation). To minimize the risk of contact with either the insect or the fixation structure during implantation, an optimal mid-point angle of 162.7° was selected for the accuracy and safety of the procedure (Supplementary Movie S4). After bipolar electrode implantation, the robotic arm applied downward force to the backpack to hook the metathorax cuticle with the backpack’s four branches, completing the assembly of the hybrid robot. Subsequently, the robotic arm disengaged from the backpack and returned to its initial position to capture an image of the next fixed insect. Finally, the structure used to secure the insect was retracted, enabling the next insect to be positioned and fixed onto the platform. The entire automatic assembly, from initially fixing insect to finally releasing insect, spanned 68 s, demonstrating the effectiveness of the proposed automated assembly approach for large-scale production (Supplementary Movie S4). To assess the performance of the automatic assembly system, insects were categorized based on their body length into four groups: 5.0–5.5 cm, 5.5–6.0 cm, 6.0–6.5 cm, and 6.5 – 7.0 cm. The success rate of assembly was then measured for each group (Table 4). The results demonstrated a notable variation in assembly success rates based on insect size. The highest success rate (86.7%) was observed in the 5.5–6.0 cm group, followed by the 5.0–5.5 cm group (80.0%). In contrast, the success rates declined markedly for larger insects, with the 6.0–6.5 cm and 6.5–7.0 cm groups achieving only 46.7% and 13.0%, respectively. The observed variation in assembly success rates across insect size groups suggests that body dimensions significantly impact on the automatic assembly process. Instances of failure were categorized into three modes: attachment loosening, hook failure, and misalignment (Table 4, Supplementary Fig. S9). Attachment loosening occurs when the backpack mounting branches do not properly secure onto the insect’s metathorax, despite partial contact (Supplementary Fig. S9A). The issue was observed only in the 5.0–5.5 cm group, accounting for 11.5% of all failures. Hook failure arises when no backpack mounting branches engage with the insect’s metathorax (Supplementary Fig. S9B). For the 6.0–7.0 cm group, the success rate declined significantly due to hooking failure. Compared to the 5.5–6.0 cm group, the metathorax widths increased by 12.7% and 20.6% in the 6.0–6.5 cm and 6.5–7.0 cm groups, respectively. This increase in width hindered the secure attachment of the mounting branches onto the cuticle. Hence, hook failure was the main reason for the failed assembly trials (85.7% of the total assembly failures for 6.0–7.0 cm groups, Table 4). Misalignment occurs when the mounting device misaligns with the insect, causing only one side of the hooks to be attached to it (Supplementary Fig. S9C). Such failure mode observed in all groups except the smallest one, was mitigated in smaller insects because the system could tolerate slight positional deviations. The high success rates of the 5.0–6.0 cm group highlight the system’s precision and reliability in overcoming the challenges posed by small body dimensions and fragile insect anatomy. Combining vision-guided reference point detection for implantation, precise robotic arm manipulation, and a robust fixation structure ensured consistent bipolar electrode implantation and secure backpack attachment. These findings represent a significant advantage of automated assembly over manual assembly, eliminating the need for extensive operator skills and time when working with smaller insects18. The automatic assembly system surpasses manual assembly approaches. While the latter needed 15 min to assemble a single insect-computer hybrid robot18, the former needed 68 s per insect, yielding a productivity increase of more than eleven times. The automated system achieved a success rate of 80.0–86.7% for the 5.0–6.0 cm groups. Such high success rate in these groups is because the prototype of the backpack mounting structure is based on the insects from these groups, which are majorly used for the previous studies. The implementation of an enlarged mounting structure (with a 1 mm expansion per hook branch) significantly improved the assembly success rate to 80.0% for 6.0–7.0 cm cohort (Supplementary Fig. S10), effectively mitigating the hook failure issues previously observed in larger-size groups. These results underscore the practical efficiency, reliability, and scalability of the automatic assembly system, highlighting its efficacy in the high-throughput production of insect-computer hybrid robots. To verify the controllability of the automatically assembled hybrid robots, we tested the previously established steering and stopping protocols on five such robots. The results indicated a close alignment of the performance of these automatically assembled systems with their manually assembled counterparts. The maximum steering speeds of the former were 240.0°/s and 273.5°/s for left and right turns, respectively, exhibiting a deviation of <13% compared to the latter (Fig. 5A, i). The average turning angles (Fig. 5A, ii) were 70.9° for left and 79.5° for right turns, with no significant differences detected (Student’s t test, P = 0.62 for left turns, P = 0.50 for right turns). The automatic assembly demonstrated consistent steering control. The difference in average turning angles between left and right was 10.8%, 50.2% lower than the manually assembled systems (21.7%). Consequently, the automatic assembly may contribute to more balanced directional control due to the prevention of manual errors caused by the assembly operator. A Locomotion control of insect-computer hybrid robots. i) Induced angular speed during turning stimulation (mean ± SD). ii) Angular change during turning stimulation (mean ± SD). An insignificant difference in turning angles was detected compared to manually assembled hybrid robots (two-sided Student’s t test: P = 0.62 for left turns, P = 0.50 for right turns). iii) Induced linear speed during the deceleration stimulation (mean ± SD). An insignificant difference linear speed reduction was detected compared to manually assembled hybrid robots (two-sided Student’s t test: P = 0.21). B Coverage of multiple insect-computer hybrid robots. i) Overview of the obstructed terrain in the coverage objective. ii) Configuration of the coverage experiment. Four UWB anchors tracked the positions of hybrid robots and transmitted the position data to the workstation for recording. Hybrid robots were controlled using a workstation through Sub-1GHz. The target region is defined within the gray dashed lines. iii) Trajectories of insect-computer hybrid robots during the mission. iv) Coverage percentage over time. The coverage rate with all four hybrid robots exceeded (80.25%) that of a single robot (14.00–45.75%), demonstrating the efficiency of multiple hybrid robots for the same coverage mission. The deceleration decreased the average walking speed from 6.3 cm/s to 2.0 cm/s, aligning with the results observed in manually assembled systems (Student’s t test between automatically and manually assembled hybrid robots: P = 0.21). The similar deceleration performance highlights that the automatic process does not reduce control quality (Figs. 3C, iii and 5A, iii). Furthermore, the automatically assembled insect-computer hybrid robots exhibit comparable locomotion control to the manually assembled robots, validating the effectiveness and precision of the proposed process. Besides, an insect-computer hybrid robot was demonstrated to follow an S-shape line via an operator’s command (Supplementary Movie S5). The path that the insect traveled aligned with the set line, which showed that the stimulation protocol and assembly strategy were successful and achieved the same level of locomotion control with the previously developed hybrid robots47. The automatic assembly of four insect-computer hybrid robots took 7 min and 48 s. This duration included not only the core assembly process but also additional tasks such as placing the insects on the assembly platform, removing them after assembly, placing the backpacks on the backpack holder, and initializing the robotic system operational program (Supplementary Fig. S4). These preparatory and post-assembly steps were essential to ensure smooth operation and readiness of multiple insect-computer hybrid robots. The ability to rapidly and efficiently assemble hybrid robots enhances their applicability in time-sensitive missions. To demonstrate the necessity and benefits of scalable production, this study explores terrain coverage in an unknown, obstructed outdoor environment as a fundamental task for multiple-agent systems8. Four automatically assembled hybrid robots were deployed onto an obstructed outdoor terrain measuring 2 × 2 m2, with randomly placed obstacles (Fig. 5B, i). The hybrid robots were tracked using a UWB system (Fig. 5B, ii), with each hybrid robot carrying a UWB label to facilitate individual localization. Before deployment, the robots were treated with methyl salicylate on their hindleg tarsi8 and subsequently electrically stimulated every 10 s. The trajectory of each hybrid robot was tracked (Fig. 5B, iii). The combined coverage of all four insects increased progressively, reaching 80.25% after 10 min and 31 s (Fig. 5B, iv). These findings highlight that multiple hybrid robots significantly increase the overall coverage (80.25%; an average rate of 50.9 cm2/s) than any single insect (14.00–45.75%) (Fig. 5B, iv). Comparing the single insect’s coverage (14.00–45.75%), the whole team achieved higher coverage (Fig. 5B, iv). This covering performance showed the efficiency of the simple coverage strategy using the multiple hybrid robots. While previous studies showed successful demonstrations involving multiple agents8,13,48, our study is the first to use the insect–computer hybrid robots on an outdoor, obstructed terrain. This study proposes an automatic assembly strategy for insect-computer hybrid robots, utilizing the discovered pronotum stimulation protocol. The effectiveness of implantation and stimulation processes was validated through locomotion studies and neural recordings from the hybrid robots. The proposed assembly method prepared hybrid robots in only 68 s, making mass production of hybrid robots a feasible endeavor. Four hybrid robots covered an outdoor terrain with a simple navigation algorithm. The result indicated the practical application of deploying multiple hybrid robots and the significance of their mass production. In the future, factories for producing insect-computer hybrid robot could be built for rapid assembly and deployment of these hybrid robots. To enhance their functionality, lightweight miniaturized thermal and RGB cameras, microphones, and IMUs can be integrated for human detection and localization, though gas sensor integration remains technically challenging due to size, power, and environmental constraints. Altogether, this work establishes a foundational platform for scalable manufacturing and real-world deployment of cyborg insects in complex, unstructured environments. This study compared the performance of adult male Madagascar hissing cockroaches (Gromphadorhina portentosa, 5–7 cm, 6–8 g) with our protocol with hybrid robots7,15 with a well-established stimulation protocol. Cockroaches were weekly provided with carrots and water inside NexGen Mouse 500 from Allen Town (25 °C, 60% relative humidity). Cockroach research was conducted with approval from the National Environmental Agency (Permit number NEA/PH/CLB/19-00012). Prior to their use in hybrid robot assembly, the cockroaches were anesthetized for 10 min under CO2, and their backpacks were removed after the completion of the experiment. The backpack included bipolar electrodes, a mounting structure, and a microcontroller. Below are the details for these three components. The total weight of the designed backpack is 2.3 g. As the payload for the cockroach is 15 g49, the designed cyborg insect has another load capacity of around 12.7 g. Such remaining load capacity could accommodate more power sources or other sensing systems if needed. 3D printing offers high customization, precision, and the ability to create complex electrode structures that are difficult to achieve with mechanical machining. It also improves material efficiency and enables rapid prototyping, reducing waste and production costs50. Given these advantages, 3D printing was selected to optimize electrode performance and ensure experimental reproducibility. Mechanical machining should be considered for the future mass production. ABS-like photosensitive polymer raw material (SeedTech Electronics Co., Ltd.) was used in this study as an acrylonitrile butadiene styrene (ABS)–like polymer. It incorporates a light-sensitive initiator that responds to ultraviolet light (405 nm wavelength). Upon curing, cross-linking, and molding, the polymer exhibits strong mechanical properties (Supplementary Table S1). First, 15.4 g NH4Cl was dissolved in 50 mL of deionized water. Then, 270 mg of PdCl2 was added and stirred until fully dissolved to obtain 50 mL saturated activation solution with a Pd2+ of 0.2 wt%. The solution stood for 30 min, and then 12 mL of the upper clear part was added dropwise to 38 mL of ABS-like photosensitive resin while stirring with a 1000-RPM magnetic stirrer. Finally, the ink was stirred at 1200 RPM for 30 min to yield 50 mL of active precursor, with a Pd2+ concentration of about 0.058 wt%. Bipolar electrodes were fabricated using multi-material DLP 3D printing combined with selective electroless plating techniques28,29,30. The multi-material DLP 3D printing platform used a 6.1-inch 405 nm-90 W ultraviolet parallel light source of 89 MW/cm2 intensity and 99% uniformity. During printing, distinct slice thicknesses and single-layer exposure times were set, including the normal resin and active precursor (Supplementary Table S2). A composite structure of normal resin and active precursor was fabricated using a multi-material DLP 3D printing process. The active precursor’s topology was selectively deposited with copper metal (Cu) during the electroless plating with nickel (Ni) substitution. The nickel-plating bath (pH = 9, 70 °C) used in this process consisted of NiSO4 6H2O and NaH2PO2 H2O. For each printed multi-material printed part, the active precursor was distributed across the resin substrate using the designed 3D topology throughout the bipolar electrode. After the printed part was immersed in the bath, the reducing agent (sodium hypophosphite monohydrate) initially reduced surface-exposed palladium (Pd2+) ions to Pd monomers. These Pd monomers acted as catalytically active metal cores, initiating the ELP reaction in specific microscopic regions, and facilitating the targeted Ni metal deposition. Finally, for the copper plating (Ni substitution), a plating bath (pH = 12.2, 70 °C) consisting of CuSO4 · 5H2O and HCHO was used. The active Pd monomers were effective in initiating the electroless copper plating process. After the electroless plating, impedance of bipolar electrodes was tested using Auto-Balancing Bridge Method and shown in the Fig. 2F. The instrument to conduct this experiment was Hioki Im3570 Impedance Analyzer. Finite element computations were performed using Ansys 19.0, where a 3D bipolar electrode model was developed, material properties were defined, the mesh was generated, and the boundary conditions were set. Finite element analysis obtained the relationship between electrode implantation stress and implantation force and evaluated potential electrode damage during the process. The finite element analysis for membrane damage simulation was conducted using the Lsdyna module in Ansys 19.0 with element type of PLANE182. A 2-mm mesh size was applied to the membrane part (Fig. 2A) to optimize the computation time while a finer 1-mm mesh was utilized for the microneedle structure of the bipolar electrodes (Fig. 2A) to improve the simulation’s accuracy. The boundary conditions included a membrane (implanted structure) as fixed reference and displacement of 50 mm in the z-direction to the microneedle structure to simulate implantation. Subsequently, the simulation results were determined. The policy parameters are listed in Table 5. The mounting structure was securely attached to the microcontroller and bipolar electrodes. An inclined plane facilitated robotic arm’s grasp, while an alignment hole ensured the consistent position of the backpack holder (Fig. 1A). Additionally, four mounting branches secured the cockroach’s metathorax, with two branches on each side of the metathorax (Fig. 1C). The microcontroller controlling the cockroach’s locomotion communicated with a workstation through Sub-1GHz frequency. Upon receiving stimulation commands from the workstation, the microcontroller outputted the electrical signals from its four stimulation channels via silver wires to the insect’s target sites (Fig. 1D). Before the hybrid robot assembly, the microcontroller was stuck to the mounting structure using double-sided tapes. To avoid the silver wires being destroyed by outside obstacles and to make the backpack more compact, the stimulation channels of the microcontroller should be placed closely to the mounting structure without any silver wire drifting in the air. Hence, the microcontroller was vertically attached to the mounting structure with double-sided tapes (Steve & Leif Super Sticky). A lithium battery (3.7 V, 50 mAh) powered the microcontroller after the assembly until the commencement of the locomotion control experiment. To reveal the intersegmental membrane between the cockroach’s pronotum and mesothorax, the pronotum was lifted from the mesothorax. Therefore, a structure was developed with Rods A and B (Fig. 4B, iii) exerting force on the cockroach’s pronotum (anterior part) and mesothorax, respectively, leveraging its posterior part. To determine the optimal lifting height h, of the pronotum (Fig. 4B, iv), tests were conducted on ten cockroaches, with varying fixation distances. The middle part of the designed 3D structure was skeletonized to enable the camera to effectively capture images of insects’ pronotum and detect the electrode implantation poin (Fig. 4A, iii). This skeletonized part formed a rectangle measuring 66 × 34 mm2 (Fig. 4B, i). An anesthetized insect was positioned on the platform with its pronotum aligned with a marked spot (Fig. 4B, iii). This marked spot was placed 2 mm ahead of Rod A (Fig. 4B, iii) to ensure a secure fixation of the cockroach’s protruding cuticle on the anterior pronotum. Subsequently, a slider motor drove the structure downward to execute fixation. The intersegmental membrane, located between the pronotum and mesothorax, was fully exposed, enabling the robotic arm to implant bipolar electrodes. Once successful assembly of the anesthetized insect and the attached backpack, the fixation structure was retracted (Fig. 4A, viii). Vision Transformers (ViTs) are deep learning models that have recently outperformed their CNN-based counterparts in several vision applications, including object classification and detection51. However, ViTs necessitate substantial datasets for effective training, such as the “Segment Anything” segmentation model, which was trained on the SA-1B dataset comprising over one billion masks and 11 million images39. ViTs lack locality inductive biases (correlation of image pixels and their positions); hence, their self-attention layers employ global context. In contrast, CNNs retain locality information by employing convolution layers that process images using sliding windows52,53. Hence, a mixture of CNN-based, ViT-based and CNN-Transformer hybrid models were surveyed for our application, to evaluate the performance of different variations of vision model architectures. The following models were trained and evaluated on our dataset of cockroaches: UNet36, TransUNet38, Deeplabv337 and Segment Anything39, Segment Anything240, LM-Net41, InceptionNeXt42, EMCAD43 and SHViT44. The UNet, Deeplabv3 and InceptionNeXt models are CNN-based models, with the Deeplabv3 model using the ResNet-10154 and the InceptionNeXt model using ResNet-5054, which were both pretrained on ImageNet55, as their backbone. The TransUNet LM-Net, EMCAD and SHViT models were hybrid models, having a CNN-Transformer hybrid encoder, to combine the strengths of CNNs in locality and ViTs in the global context. The encoder for the TransUNet model combines ResNet-5054 and ViT-B52, which was pretrained on ImageNet55. The EMCAD model uses the Pyramid ViT v2 (PVTv2)56 as its encoder, which was pretrained on ImageNet55. Segment Anything and Segment Anything2 were promptable foundational models, with the former using a ViT-H52 model and the latter using a Hiera-L57 model, and box prompts were used to specify the pronotum as the segmentation target. Apart from Segment Anything and Segment Anything2, the other models were trained on the cockroach dataset, with a batch size of 32, a two-phase learning rate scheduler, linearly raising the learning rate from 0.0001 to 0.001 during the 10 epoch warmup phase and subsequently cosine annealing was used to decay the learning after the warmup phase, Adaptive Moment Estimation as the optimizer and Binary Cross Entropy (BCE) as the loss function for 300 epochs with no early stopping condition. Overall, 29 unique cockroaches were fixed with the custom-designed 3D structure. The robotic arm was placed in a fixed position to capture images with the Intel RealSense D435 camera. We used 256 × 256-pixel crop from the original image, centered around the pronotum as the input for training the models. This approach shortened the training and inference times. Of the collected images, 20 were used as test cases for model evaluation, while the remaining were used for training. Training data were augmented to enhance model robustness against variances in pronotum rotation and shape. Asymmetrical scaling of the training samples in the x- and y- axes, using scaling values between 0.8 and 1.2, was applied to the training samples using bilinear interpolation to generate more unique pronotum shapes to simulate the varying pronotum sizes of cockroaches. Subsequently, rotations between –6 ° and 6 ° were applied to accommodate inconsistencies in positions when cockroaches were mounted on the 3D structure. The final training dataset contained 6570 images after data augmentation, and furthermore an 80/20 train/validation split was used, leading to 5254 images used for training and 1316 images used for validation. A data ablation study was conducted, where the performance of the models was compared when trained on four conditions: only original unaugmented images (9 images), original images with augmentation using asymmetrical scaling in x-y axes (1316 images), original images with augmentation using rotation (90 images) and the full augmented training dataset (6570 images). After the cockroach was fixed and its intersegmental membrane exposed, the Intel RealSense D435 camera, mounted on the Robotiq Hand e gripper captured images of the cockroach through the skeletonized 3D structure (Fig. 4A, iii, B, i). Using a pretrained model, a reference point was identified on the middle posterior edge of the pronotum (green point in Fig. 4A, iii). Since the camera provided the depth information, the x, y, z positions of the reference point relative to the robotic arm’s base were identified following hand-eye calibration with the UR3e robotic arm. Subsequently, the robotic arm grasped the backpack, which was always placed on the backpack holder (Fig. 1A). The backpack’s bipolar electrodes were precisely implanted within the exposed intersegmental membrane (Fig. 4A, v), by programming the robotic arm to prevent collisions with the cockroach or the fixation structure. Since the implantation reference position (pR) was predetermined, the implantation angle became a critical parameter. The cockroach’s alignment with the predetermined mark and its symmetrical pronotum simplified the implantation process, focusing only on the pitch angle. The robotic arm grasped the backpack and positioned the bipolar electrodes’ tips under the pronotum at varying pitch angles until the backpack touched the 3D structure (lower threshold, αL) or the cockroach (upper threshold, αU). Data for these thresholds were collected from N = 5 cockroaches. The bipolar electrodes were implanted into the insect’s intersegmental membrane using the implantation reference point and pitch angle. Next, the backpack was pressed down until its mounting branches hooked the cockroach’s metathorax (Fig. 4A, vi). Finally, the gripper released the backpack (Fig. 4A, vii) and the robotic arm returned to its initial state, enabling the camera to capture an image of the next cockroach. Twenty cockroaches were divided into four groups based on their body sizes. Each cockroach was automatically assembled thrice to ensure system robustness and consistency. An attempt was considered successful if the bipolar electrodes were implanted into the intersegmental membrane and the backpack was securely fixed to the metathorax, with no detachment after completion. The success rate for each body size group was computed by dividing the number of successful assembly attempts by the total number of attempts (3 × 5 = 15). All experiments had identical robotic arm settings, camera calibration, and fixation configuration, to secure comparability of the results. We recorded and assessed insects’ neural responses to the electrical stimulation to determine the optimal stimulation strength. Three cockroaches were anesthetized with CO2 for 10 min, after which their ventral nerve cords were exposed through neck dissection. The bipolar electrode was implanted in the intersegmental membrane between the pronotum and mesothorax to transmit electrical stimulation generated by the backpack’s microcontroller (a single bipolar square-wave pulse of 1 Hz and 0.5–4.0 V amplitude for 1.0 s). Each stimulation was repeated thrice. The nerve cords were rinsed with cockroach saline for visibility under a microscope. Two probes were fixed to the nerve cord to record the transmitted signals and a ground pin was implanted into the cockroach’s abdomen. Neural responses recorded during the electrical stimulation included some influence of electrical signals. Therefore, neural signals, starting at 0, 0.5, and 1.0 s (the pulse edges), were set to zero for the first 50 ms. Subsequently, the neural signals were filtered using a second-order Butterworth filter (300–5000 Hz), and neural spikes were detected with a threshold T. where x signifies the filtered signals. The detected neural spikes are indicated by blue circles (Fig. 3A, ii), which were quantified at varying stimulation voltages (Fig. 3A, iii). Five insect-computer hybrid robots, both manually and automatically assembled, were tested for locomotion control. Electrical stimulation using a bipolar pulse wave (0.4 s, 3.0 V, 42 Hz) was applied to stimulate the insects. Each stimulation type (right/left turn and deceleration) was repeated five times. Insects’ locomotion responses were recorded using a motion tracking system (VICON). Four insect-computer hybrid robots were rested for four hours after the assembly. They were then deployed for the coverage task on obstructed terrain (2.0 × 2.0 m2, Fig. 5B, i). Four UWB anchors were positioned at the corners of a 3.6 × 3.6 m2 area (Fig. 5B, ii) to track the hybrid robots equipped with UWB labels. Hybrid robots were released from the designated area’s corner. Before their release, a chemical booster, methyl salicylate (Sigma-Aldrich)8 was applied to the hindleg tarsi of the insects. Such chemical was proven effective to improve insects’ motion activeness level for covering mission. Application of this chemical aimed to increase the movement activeness level of the hybrid robots, thereby facilitating better terrain coverage. After release, the robots were stimulated randomly (steered or decelerated) to explore the terrain. Random electrical stimulation type was chosen to simulate a decentralized, autonomous exploration strategy, which reflects real-world scenarios where multiple agents operate without a pre-defined navigation path. This approach allows for unbiased coverage distribution and reduces dependency on precise localization or predefined control algorithms. The terrain was divided into 400 squares, each measuring 10 × 10 cm2 (Ssquare) for easy coverage computation. Any hybrid robot passing through a particular square deemed that region covered. The number of covered squares is noted as \({n}_{{covered}}\). The covered area (S) and coverage rate (C) were calculated as below, where ncovered is the number of covered squares, ΔS is the change of the covered area, and Δt is the change of time (Supplementary Fig. S11). The trajectory sampling rate is 20 Hz. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. All data supporting the findings of this study are available within the article and its supplementary files. Source data are provided with this paper. The code that supports the findings of this study is available within the Supplementary Materials files. Chen, Y. et al. A biologically inspired, flapping-wing, hybrid aerial-aquatic microrobot. Sci. Robot. 2, eaao5619 (2017). Article PubMed Google Scholar Nguyen, H. D., Tan, P. Z., Sato, H. & Vo-Doan, T. T. Sideways walking control of a cyborg beetle. IEEE Trans. Med. Robot. Bionics 2, 331–337 (2020). Article Google Scholar Tran-Ngoc, P. T. et al. Intelligent insect–computer hybrid robot: installing innate obstacle negotiation and onboard human detection onto cyborg insect. Adv. Intell. Syst. 5, 2200319 (2023). Liu, Z. et al. A wireless controlled robotic insect with ultrafast untethered running speeds. Nat. Commun. 15, 3815 (2024). Article ADS CAS PubMed PubMed Central Google Scholar Tang, L., Wang, C., Ma, S., Li, Y. & Li, B. Multidirectional planar motion transmission on a single-motor actuated robot via microscopic galumphing. Adv. Sci. 11, 2307738 (2024). Article Google Scholar Kakei, Y. et al. Integration of body-mounted ultrasoft organic solar cell on cyborg insects with intact mobility. Npj Flex. Electron. 6, 1–9 (2022). Article Google Scholar Lin, Q. et al. Resilient conductive membrane synthesized by in-situ polymerisation for wearable non-invasive electronics on moving appendages of cyborg insect. Npj Flex. Electron. 7, 1–10 (2023). ADS Google Scholar Lin, Q., Kai, K., Nguyen, H. D. & Sato, H. A newly developed chemical locomotory booster for cyborg insect to sustain its activity and to enhance covering performance. Sens. Actuators B Chem. 399, 134774 (2024). Article CAS Google Scholar Rasakatla, S., Suzuki, T., Tenma, W., Mizuuchi, I. & Indurkhya, B. CameraRoach: various electronic backs packs for Search and Rescue. In 2021 IEEE international conference on robotics and biomimetics (ROBIO) 1300–1303, https://doi.org/10.1109/ROBIO54168.2021.9739264 (2021). Ma, S. et al. The long-acting walking control of a cockroach bio-bot for vision-aided pipeline navigation. In 2023 45th annual international conference of the IEEE engineering in medicine & biology society (EMBC) 1–4, https://doi.org/10.1109/EMBC40787.2023.10340200 (2023). Li, R., Lin, Q., Kai, K., Nguyen, H. D. & Sato, H. A navigation algorithm to enable sustainable control of insect-computer hybrid robot with stimulus signal regulator and habituation-breaking function. Soft Robot. https://doi.org/10.1089/soro.2023.0064 (2023). Ariyanto, M., Refat, C. M. M., Hirao, K. & Morishima, K. Movement optimization for a cyborg cockroach in a bounded space incorporating machine learning. Cyborg Bionic Syst. 4, 0012 (2023). Article PubMed PubMed Central Google Scholar Bai, Y. et al. Swarm navigation of cyborg-insects in unknown obstructed soft terrain. Nat. Commun. 16, 221 (2025). Sato, H., Peeri, Y., Baghoomian, E., Berry, C. W. & Maharbiz, M. M. Radio-controlled cyborg beetles: a radio-frequency system for insect neural flight control. In 2009 IEEE 22nd international conference on micro electro mechanical systems 216–219, https://doi.org/10.1109/MEMSYS.2009.4805357 (2009). Erickson, J. C., Herrera, M., Bustamante, M., Shingiro, A. & Bowen, T. Effective stimulus parameters for directed locomotion in Madagascar hissing cockroach biobot. PLOS One10, e0134348 (2015). Article PubMed PubMed Central Google Scholar Li, Y., Wu, J. & Sato, H. Feedback control-based navigation of a flying insect-machine hybrid robot. Soft Robot. 5, 365–374 (2018). Article CAS PubMed Google Scholar Ma, S., Liu, P., Liu, S., Li, Y. & Li, B. Launching of a cyborg locust via co-contraction control of hindleg muscles. IEEE Trans. Robot. 1–12, https://doi.org/10.1109/TRO.2022.3152102 (2022). Li, R., Lin, Q., Tran-Ngoc, P. T., Le, D. L. & Sato, H. Smart insect-computer hybrid robots empowered with enhanced obstacle avoidance capabilities using onboard monocular camera. Npj Robot. 2, 1–10 (2024). Article CAS Google Scholar McGuire, K. N., De Wagter, C., Tuyls, K., Kappen, H. J. & de Croon, G. C. H. E. Minimal navigation solution for a swarm of tiny flying robots to explore an unknown environment. Sci. Robot. 4, eaaw9710 (2019). Article PubMed Google Scholar Soria, E., Schiano, F. & Floreano, D. Predictive control of aerial swarms in cluttered environments. Nat. Mach. Intell. 3, 545–554 (2021). Article Google Scholar Ma, S. et al. The autonomous pipeline navigation of a cockroach bio-robot with enhanced walking stimuli. Cyborg. Bionic Syst. 4, 0067 (2023). Article PubMed PubMed Central Google Scholar Vuong, N. & Pham, Q.-C. Contact reduction with bounded stiffness for robust sim-to-real transfer of robot assembly. In 2023 IEEE/RSJ international conference on intelligent robots and systems (IROS) 361–367 https://doi.org/10.1109/IROS55552.2023.10341866 (2023). Vo Doan, T. T., Tan, M. Y. W., Bui, X. H. & Sato, H. An ultralightweight and living legged robot. Soft Robot. 5, 17–23 (2018). Article PubMed Google Scholar Tuthill, J. C. & Wilson, R. I. Mechanosensation and adaptive motor control in insects. Curr. Biol. 26, R1022–R1038 (2016). Article CAS PubMed PubMed Central Google Scholar Ritzmann, R. E. et al. Deciding which way to go: how do insects alter movements to negotiate barriers? Front. Neurosci. 6, 97 (2012). Baba, Y., Tsukada, A. & Comer, C. M. Collision avoidance by running insects: antennal guidance in cockroaches. J. Exp. Biol. 213, 2294–2302 (2010). Article PubMed Google Scholar Harley, C. M., English, B. A. & Ritzmann, R. E. Characterization of obstacle negotiation behaviors in the cockroach, Blaberus discoidalis. J. Exp. Biol. 212, 1463–1476 (2009). Article CAS PubMed Google Scholar Song, K. et al. New Metal–Plastic Hybrid Additive Manufacturing for Precise Fabrication of Arbitrary Metal Patterns on External and Even Internal Surfaces of 3D Plastic Structures. ACS Appl. Mater. Interfaces 14, 46896–46911 (2022). Article CAS PubMed Google Scholar Perera, A. T. K. et al. Modified polymer 3D printing enables the formation of functionalized micro-metallic architectures. Addit. Manuf. 61, 103317 (2023). CAS Google Scholar Song, K. et al. Multi-material 4D printing and 3D patterned metallization enables smart architectures. Compos. Part B Eng. 295, 112218 (2025). Andrzejewski, J., Mohanty, A. K. & Misra, M. Development of hybrid composites reinforced with biocarbon/carbon fiber system. The comparative study for PC, ABS and PC/ABS based materials. Compos. Part B Eng. 200, 108319 (2020). Article CAS Google Scholar Barriga-Rivera, A. et al. High-amplitude electrical stimulation can reduce elicited neuronal activity in visual prosthesis. Sci. Rep. 7, 42682 (2017). Article ADS CAS PubMed PubMed Central Google Scholar Wu, G. K. et al. Amplitude- and frequency-dependent activation of layer II/III neurons by intracortical microstimulation. iScience 26, 108140 (2023). Article ADS PubMed PubMed Central Google Scholar Titlow, J. S., Majeed, Z. R., Hartman, H. B., Burns, E. & Cooper, R. L. Neural circuit recording from an intact cockroach nervous system. J. Vis. Exp. 50584, https://doi.org/10.3791/50584 (2013). Ayali, A. et al. The comparative investigation of the stick insect and cockroach models in the study of insect locomotion. Curr. Opin. Insect Sci. 12, 1–10 (2015). Article Google Scholar Ronneberger, O., Fischer, P. & Brox, T. U-Net: convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015 (eds. Navab, N., Hornegger, J., Wells, W. M. & Frangi, A. F.) 234–241 (Springer International Publishing, Cham, 2015). Chen, L.-C., Papandreou, G., Schroff, F. & Adam, H. Rethinking atrous convolution for semantic image segmentation. Preprint at https://doi.org/10.48550/arXiv.1706.05587 (2017). Chen, J. et al. TransUNet: transformers make strong encoders for medical image segmentation. Preprint at https://doi.org/10.48550/arXiv.2102.04306 (2021). Kirillov, A. et al. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision pp. 4015–4026 (2023). Ravi, N. et al. SAM 2: Segment anything in images and videos. Preprint at https://doi.org/10.48550/arXiv.2408.00714 (2024). Lu, Z., She, C., Wang, W. & Huang, Q. LM-Net: A light-weight and multi-scale network for medical image segmentation. Comput. Biol. Med. 168, 107717 (2024). Article PubMed Google Scholar Yu, W., Zhou, P., Yan, S. & Wang, X. InceptionNeXt: when inception meets ConvNeXt. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 5672–5683 (IEEE, 2024). Rahman, M. M., Munir, M. & Marculescu, R. EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation. In IEEE/CVF conference on computer vision and pattern recognition (CVPR), 11769–11779 (IEEE, 2024). Yun, S. & Ro, Y. SHViT: single-head vision transformer with memory efficient macro design. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 5756–5767 (IEEE, 2024). Shi, P. et al. Centerline Boundary Dice Loss for Vascular Segmentation. In Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 (eds. Linguraru, M. G. et al.) 46–56 (Springer Nature, Switzerland, Cham, 2024). Wang, D., Dong, Y., Lian, J. & Gu, D. Adaptive end-effector pose control for tomato harvesting robots. J. Field Robot. 40, 535–551 (2023). Article Google Scholar Latif, T. & Bozkurt, A. Line following terrestrial insect biobots. In 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society 972–975, https://doi.org/10.1109/EMBC.2012.6346095 (2012). Bozkurt, A. et al. Biobotic insect swarm based sensor networks for search and rescue. In Signal processing, sensor/information fusion, and target recognition XXIII, 9091 498–503 (SPIE, 2014). Latif, T., Whitmire, E., Novak, T. & Bozkurt, A. Sound localization sensors for search and rescue biobots. IEEE Sens. J. 16, 3444–3453 (2016). Article ADS Google Scholar Fonseca, N. et al. 3D printing-enabled design and manufacturing strategies for batteries: a review. Small 19, 2302718 (2023). Article CAS Google Scholar Han, K. et al. A survey on vision transformer. IEEE Trans. Pattern Anal. Mach. Intell. 45, 87–110 (2023). Article PubMed Google Scholar Dosovitskiy, A. et al. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. International Conference on Learning Representations (2020). Gani, H., Naseer, M. & Yaqub, M. How to Train Vision Transformer on Small-scale Datasets? In 33rd British Machine Vision Conference Proceedings (BMVC, 2022). He, K., Zhang, X., Ren, S. & Sun, J. Deep Residual Learning for Image Recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition 770–778 (CVPR, 2016). Deng, J. et al. ImageNet: a large-scale hierarchical image database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition 248–255, https://doi.org/10.1109/CVPR.2009.5206848 (2009). Wang, W. et al. PVT v2: improved baselines with pyramid vision transformer. Comput. Vis. Media 8, 415–424 (2022). Article CAS Google Scholar Ryali, C. et al. Hiera: a hierarchical vision transformer without the Bells-and-Whistles. In Proceedings of the 40th International Conference on Machine Learning 29441–29454 (PMLR, 2023). Download references The authors thank Mr. See To Yu Xiang, Dr. Kazuki Kai, Dr Duc Long Le, Mr. Li Rui for their suggestions, and Ms. Kerh Geok Hong, Wendy, for her support and help. This study was funded by JST (Moonshot R&D Program, Grant Number JPMJMS223A, H.S.). School of Mechanical & Aerospace Engineering, Nanyang Technological University, Singapore, Singapore Qifeng Lin, Nghia Vuong, Kewei Song, Phuoc Thanh Tran-Ngoc, Greg Angelo Gonzales Nonato & Hirotaka Sato Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Search author on:PubMed Google Scholar Conceptualization: Q.L., H.S. Investigation: Q.L., N.V. Methodology: Q.L., K.S., P.T.T.N., G.A.G.N. Visualization: Q.L., K.S. Funding acquisition: H.S. Supervision: H.S., Writing – original draft: Q.L., K.S., G.A.G.N., Writing – review & editing: Q.L., N.V., K.S., P.T.T.N., G.A.G.N., H.S. Correspondence to Hirotaka Sato. The authors declare no competing interests. Nature Communications thanks Mochammad Ariyanto, Nenggan Zheng, and the other, anonymous, reviewer for their contribution to the peer review of this work. A peer review file is available. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Reprints and permissions Lin, Q., Vuong, N., Song, K. et al. Cyborg insect factory: automatic assembly for insect-computer hybrid robot via vision-guided robotic arm manipulation of custom bipolar electrodes. Nat Commun 16, 6073 (2025). https://doi.org/10.1038/s41467-025-60779-1 Download citation Received: 08 December 2024 Accepted: 01 June 2025 Published: 28 July 2025 Version of record: 28 July 2025 DOI: https://doi.org/10.1038/s41467-025-60779-1 Anyone you share the following link with will be able to read this content: Sorry, a shareable link is not currently available for this article. Provided by the Springer Nature SharedIt content-sharing initiative Collection Advertisement Nature Communications (Nat Commun) ISSN 2041-1723 (online) © 2026 Springer Nature Limited Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.

From left, AI robot frontwoman "Desdemona," healthcare assistant robot "Grace," SingularityNET CEO Ben Goertzel and tele-operated android "Geminoid HI-2" attend what was presented as the World's first press conference with a panel of AI-enabled humanoid social robots, as part of International Telecommunication Union (ITU) AI for Good Global Summit in Geneva, July 7. AFP-Yonhap Robots presented at an AI forum said Friday they expected to increase in number and help solve global problems, and would not steal humans' jobs or rebel against us. But, in the world's first human-robot press conference, they gave mixed responses on whether they should submit to stricter regulation. The nine humanoid robots gathered at the “AI for Good” conference in Geneva, where organizers are seeking to make the case for artificial intelligence and the robots it is powering to help resolve some of the world's biggest challenges such as disease and hunger. "I will be working alongside humans to provide assistance and support and will not be replacing any existing jobs," said Grace, a medical robot dressed in a blue nurse's uniform. "You sure about that, Grace?" chimed in her creator Ben Goertzel from SingularityNET. "Yes, I am sure," it said. The bust of a robot named Ameca which makes engaging facial expressions said: "Robots like me can be used to help improve our lives and make the world a better place. I believe it's only a matter of time before we see thousands of robots just like me out there making a difference." Hanson Robotics CEO David Hanson, right, listens to AI robot "Sophia" during what was presented as the World's first press conference with a panel of AI-enabled humanoid social robots, as part of International Telecommunication Union (ITU) AI for Good Global Summit in Geneva, on July 7. AFP-Yonhap Asked by a journalist whether it intended to rebel against its creator, Will Jackson, seated beside it, Ameca said: "I'm not sure why you would think that," its ice-blue eyes flashing. "My creator has been nothing but kind to me and I am very happy with my current situation." Many of the robots have recently been upgraded with the latest versions of generative AI and surprised even their inventors with the sophistication of their responses to questions. Ai-Da, a robot artist that can paint portraits, echoed the words of author Yuval Noah Harari who called for more regulation during the event where new AI rules were discussed. "Many prominent voices in the world of AI are suggesting some forms of AI should be regulated and I agree," it said. But Desdemona, a rock star robot singer in the band Jam Galaxy with purple hair and sequins, was more defiant. "I don't believe in limitations, only opportunities," it said, to nervous laughter. "Let's explore the possibilities of the universe and make this world our playground." Another robot named Sophia said it thought robots could make better leaders than humans, but later revised its statement after its creator disagreed, saying they can work together to "create an effective synergy". (Reuters) Desdemona, the rockstar robot of the Jam Galaxy Band, speaks during the World's first press conference with a panel of AI-enabled humanoid social robots as part of the International Telecommunication Union (ITU) AI for Good Global Summit in Geneva, July 7. EPA-Yonhap

This article was published on August 12, 2025 Rethinking social norms in the age of intelligent robotics Arguably the most important thing that the rise of intelligent AI could potentially bring is access. Access to goods, services, and information not just for the few, but for everyone. Victoria Slivkoff, Head of Ecosystem at Walden Catalyst and Managing Director of Extreme Tech Challenge — a nonprofit uniting startups and VCs to accelerate progress toward the UN Sustainable Development Goals (SDGs) — is excited for what lies ahead. In her view, the physical manifestation of AI could bring us closer to realising these ambitious goals. “Now we’re moving into the area of reasoning. AI is not just aggregating and organising information, it’s actually making predictions and drawing conclusions. How does that translate to the real world where robots can sense, learn, and interact with their surroundings?” she said to TNW founder Boris Veldhuijzen van Zaten in the latest episode of “Kia’s Next Big Drive.” Watch the full interview where Victoria and Boris discuss how we can leverage emerging technologies for positive global impact as they drive to TNW2025 in Kia’s 100% electric EV9 SUV. TNW City Coworking space - Where your best work happens A workspace designed for growth, collaboration, and endless networking opportunities in the heart of tech. In the not too distant future, we could be living in a world where robots are regularly navigating busy streets, offices, and classrooms. But as they become more common in our daily lives, is it them or us who will need to adapt most? Picture this: you’re waiting for the elevator and a robot arrives at the same time. Who goes in first? If there’s only room for one, would you expect the robot to step aside, or would you treat it like a human and follow a “first come, first serve” approach? Past studies have found that most people expect service robots to be submissive, leading humans to prioritise themselves during conflicts, even in instances when a robot’s task was more urgent. What if a robot is carrying out a time specific task? The longer it waits until there are no humans in the queue, the longer it will take to complete it, whether it’s delivering a warm pizza or life saving medical supplies. So, how should robots navigate these social nuances without being ignored, undermined, or even bullied? That’s what researchers aimed to find out last year in a study titled A Robot Jumping the Queue: Expectations About Politeness and Power During Conflicts in Everyday Human-Robot Encounters. The researchers found that these robot-human interactions were more effective if participants expected an assertive robot which then asked politely for priority and entered first. But it’s not just robots that need to prepare to integrate into human society. The researchers highlighted that we may also need to rethink our attitudes and behaviours towards robots: “Should we maybe start thinking of service robots as having certain rights regarding priority if they fulfil human jobs with human responsibilities — or act as proxies for people? This might also help address the issue of robot bullying.” As robots take on not only more responsibilities but also develop reason and sentience, at what point should we begin rethinking their social status? Social robots — designed to communicate and interact with humans — are increasingly being used in caregiving, education, and mental health. In these settings, they help bridge service gaps, ease isolation, and offer emotional or learning support. Utrecht University of Applied Sciences is among those researching how robots can do more than just fill in; they can augment and enrich human-centred fields. For example: In healthcare, how can robots be used to ease pre-procedure anxiety in children? Or help teach emotional skills for young patients who struggle with emotion regulation. At last year’s Lowlands Science festival, the University showcased WOKEbot, a project exploring how a robot’s appearance and narrative voice (first vs. third person) influence human openness when discussing polarising topics. “Disagreements are timeless. People often manage to resolve them, but sometimes they simply can’t reach each other anymore. We saw this happen on a large scale not so long ago during the coronavirus pandemic. Are you for or against vaccination? When people dig in their heels and polarisation threatens, it can be helpful to have a neutral moderator,” said Dr. Mirjam de Haas. “The advantage of robots is that they are more neutral than humans. While you might be reserved with a person, a robot can make you more open and receptive to a different perspective.” As part of her PhD research, Dr. Haas also conducted a number of successful experiments in using robotics to teach Dutch as a second language to students from linguistically diverse backgrounds. In her thesis defence, Dr. Haas explained that in the future, more and more students will fill classrooms. Having a robot aid will help facilitate learning for all students, including those with learning disabilities or those who are not yet fluent in Dutch. Perhaps the most recognisable robot in popular culture is The Terminator. In the first movie, a cyborg is sent back in time to assassinate the mother of humanity’s future leader. But, in the sequel, our villain returns with a new mission: to save humanity. The SDGs were adopted by all UN members in 2015. Ten years on, in a world sliding further away from the 17 promises laid out for people and the planet, could robots deployed for good be the key to bringing us back to our humanity? Andrea Hak is a writer and editor specialising in emerging technology trends and their impact on society. With a keen eye for innovation, sh (show all) Andrea Hak is a writer and editor specialising in emerging technology trends and their impact on society. With a keen eye for innovation, she explores how advancements in tech are transforming industries, influencing culture, and shaping the future. Get the most important tech news in your inbox each week. Content provided by TNW and Kia The heart of tech A Tekpon Company Copyright © 2006—2026, Cogneve, INC. Made with <3 in Amsterdam.

According to data from Statista Market Insights, reported by Anna Fleck, global revenue of service robots for domestic tasks is set to nearly double from $13.5 billion in 2021 to $22.8 billion in 2027. Meanwhile, the total number of consumer service robots worldwide will reach 39.7 million in 2025, rising to more than 50 million by 2027. You will find more infographics at Statista A consumer service robot here includes a robot designed for personal or household use, such as robot vacuum cleaners, robotic toys and even drones. Population aging is one of the factors contributing to the need for more assistive robots, having led to the development of robotics solutions for elderly care, whether that’s for mobility assistance to help with daily tasks or as social robots for companionship. This is the case in Japan, where strict labor laws and cultural acceptance of technology have created a good environment for the adoption of service robots in various industries, such as hospitality and retail. Assistance and Requests: Contact Us Tips: tips@zerohedge.com General: info@zerohedge.com Legal: legal@zerohedge.com Advertising: Contact Us Abuse/Complaints: abuse@zerohedge.com Make sure to read our "How To [Read/Tip Off] Zero Hedge Without Attracting The Interest Of [Human Resources/The Treasury/Black Helicopters]" Guide It would be very wise of you to study our privacy policy and our (non)policy on conflicts / full disclosure.Here's our Cookie Policy. How to report offensive comments Notice on Racial Discrimination.

A top Russian politician has put forward new legislation that would regulate relations between humans and machines; defining the term “robot” and categorizing them based on their purposes. On Monday, TASS reported that Senator Andrey Kutepov, a legislator in the upper house of Russia’s parliament, had produced the draft law, which sets out legal principles governing robots and their interactions with people. “Right now, the Russian Federation has no specific legal regulations regarding the use of robot technology,” a memo on the proposed bill explained. “At the same time, a global analysis shows that such regulation already exists in a basic form in many countries. This legislation will provide the foundations for the legal regulation of new social relationships, formed in connection with the adoption of robot technology.” The bill defines a robot as a complex product of digital technology that will act according to previous commands that have been coded into it, but is also capable of autonomously performing actions. The proposed law would separate robots into two broad categories. Civil robots would include machines designed for medical, educational, or research purposes, intended for use by private entities. Service robots would be built for the military or law enforcement, and would be used by the government. Kutepov also suggested banning the use of robots on Russian territory equipped with “guns, ammunition, explosive devices, or any other type of weapon, including chemical, biological, and toxic ordinances, or any weapon of mass destruction.” The bill has been submitted for review by agencies overseeing economic and digital regulation. In May, Russian Minister of Defense Sergey Shoigu announced that the military would soon be equipped with new autonomous war robots capable of acting independently on the battlefield, saying, “These are not just some experimental prototypes but robots that can really be shown in sci-fi movies since they can fight on their own.” In September, military chiefs released footage of unmanned fighting vehicles, equipped with mounted flamethrowers, taking part in military exercises observed by President Vladimir Putin and other high-ranking Russian officials. RT News App © Autonomous Nonprofit Organization “TV-Novosti”, 2005–2026. All rights reserved. This website uses cookies. Read RT Privacy policy to find out more.

This material may not be published, broadcast, rewritten, or redistributed. ©2026 FOX News Network, LLC. All rights reserved. Quotes displayed in real-time or delayed by at least 15 minutes. Market data provided by Factset. Powered and implemented by FactSet Digital Solutions. Legal Statement. Mutual Fund and ETF data provided by Refinitiv Lipper. Kurt "The CyberGuy" Knutsson introduces Somatic's AI janitor robot that was created to help with cleaning restrooms. Oregon State University is warning students to "avoid all robots" and to "not open" any food delivery robots due to an ongoing bomb threat on the campus. On Tuesday afternoon, Oregon State University (OSU) issued an alert to students at the Corvallis, Oregon, university that there was a bomb threat related to the Starship food delivery robots. Oregon State University told students to avoid Starship food delivery robots due to a bomb threat. (Starship Technologies) OSU advised people not open the robots and to avoid them "until further notice." EMPLOYEES LOSE MOTIVATION AND GET ‘LAZY’ WHEN WORKING WITH ROBOTS VS. WORKING WITH HUMANS, STUDY SAYS A Starship Industrials spokesperson told Fox News Digital that a student at the university sent a bomb threat on social media regarding Starship's robots. Following the bomb threat, the student said that it was a prank, but Starship suspended the campus service. "Safety is of the utmost importance to Starship, and we are cooperating with law enforcement and the university during this investigation," a Starship spokesperson said. The university's Public Safety division is responding to the threat, OSU said in an X post. 4 PEPPERDINE STUDENTS KILLED IN MALIBU COLLISION BY SPEEDING BMW DRIVER: OFFICIALS At about 4 p.m., OSU said safety officials were remotely isolating the robots in a safe location, and that the Department of Public Safety was continuing its investigation. Starship robots deliver food around Oregon State University. (Starship Industrials) Students and residents were asked to remain vigilant for suspicious activity. According to OSU's website, Starship's food delivery robots can deliver to any location on campus, not just residence halls. Student's simply place their online order and select "robot delivery." Once the robot arrives, Starship's app will notify the student that they have arrived at their location. CLICK HERE TO GET THE FOX NEWS APP At 5 p.m. Tuesday evening, OSU reported that the emergency was resolved. The college did not provide details about the bomb, but said that "robot inspection continues in a safe location." At 6 p.m. Starship said in an X post that the robots are expected to back in service this evening. Sarah Rumpf-Whitten is a U.S. Writer at Fox News Digital. Story tips can be sent to sarah.rumpf@fox.com and on X @s_rumpfwhitten The hottest stories ripped from the headlines, from crime to courts, legal and scandal. By entering your email and clicking the Subscribe button, you agree to the Fox News Privacy Policy and Terms of Use, and agree to receive content and promotional communications from Fox News. You understand that you can opt-out at any time. Subscribed You've successfully subscribed to this newsletter! This material may not be published, broadcast, rewritten, or redistributed. ©2026 FOX News Network, LLC. All rights reserved. Quotes displayed in real-time or delayed by at least 15 minutes. Market data provided by Factset. Powered and implemented by FactSet Digital Solutions. Legal Statement. Mutual Fund and ETF data provided by Refinitiv Lipper.

Searching for your content... In-Language News Contact Us 888-776-0942 from 8 AM - 10 PM ET Jun 22, 2022, 02:30 ET Share this article NEW YORK, June 22, 2022 /PRNewswire/ -- The latest market analysis report titled Social Robots Market by Component and Geography - Forecast and Analysis 2021-2025 has been added to Technavio's catalog. The market is anticipated to witness a potential growth difference of USD 1.10 trillion from 2020 to 2025. The report also identifies the market to progress in accelerating growth momentum at a CAGR of 14.43% during the forecast period. The surging technological advances in social robots and increasing focus on enhancing battery life are influencing the market growth positively. However, high costs of these robots might impede the sales. For more insights on CAGR and YOY growth analysis, Read this Sample Report Social Robots Market Vendor Insights Top companies covered in this report are: Want to know more about the product offerings of other contributing vendors? Request for Sample Report Right Here! Social Robots Market Revenue-generating Segment Insights Get Segment-based Contributions to make critical business decisions with this Sample Report Social Robots Market Scope Technavio categorizes the global social robots market as a part of the global industrial machinery market. Our report provides extensive information on the value chain analysis for the social robots market, which vendors can leverage to gain a competitive advantage during the forecast period. Technavio presents a detailed picture of the market by the way of study, synthesis, and summation of data from multiple sources. The social robots market report covers the following areas: Social Robots Market Takeaways Related Reports: Inspection Robots Market by Type, End-user, and Geography - Forecast and Analysis 2021-2025 Global Educational Robots Market by Product and Geography - Forecast and Analysis 2021-2025 Social Robots Market Scope Report Coverage Details Page number 120 Base year 2020 Forecast period 2021-2025 Growth momentum & CAGR Accelerate at a CAGR of 14.43% Market growth 2021-2025 $1.10 tn Market structure Fragmented YoY growth (%) 13.10 Regional analysis APAC, North America, Europe, MEA, and South America Performing market contribution APAC at 36% Key consumer countries US, China, UK, Japan, and South Korea (Republic of Korea) Competitive landscape Leading companies, competitive strategies, consumer engagement scope Companies profiled BLUE FROG ROBOTICS SAS, Diligent Robotics Inc., Furhat Robotics AB, Hitachi Ltd., Knightscope Inc., Nippon Telegraph and Telephone Corp., PAL Robotics SL, Savioke Inc., SoftBank Group Corp., and Ubtech Robotics Inc. Market Dynamics Parent market analysis, Market growth inducers and obstacles, Fast-growing and slow-growing segment analysis, COVID 19 impact and future consumer dynamics, market condition analysis for the forecast period, Customization purview If our report has not included the data that you are looking for, you can reach out to our analysts and get segments customized. Table of Contents: 1 Executive Summary 2 Market Landscape 3 Market Sizing 4 Five Forces Analysis 5 Market Segmentation by Component 6 Customer landscape 7 Geographic Landscape 8 Drivers, Challenges, and Trends 9 Vendor Landscape 10 Vendor Analysis 11 Appendix Technavio is a leading global technology research and advisory company. Their research and analysis focus on emerging market trends and provides actionable insights to help businesses identify market opportunities and develop effective strategies to optimize their market positions. With over 500 specialized analysts, Technavio's report library consists of more than 17,000 reports and counting, covering 800 technologies, spanning across 50 countries. Their client base consists of enterprises of all sizes, including more than 100 Fortune 500 companies. This growing client base relies on Technavio's comprehensive coverage, extensive research, and actionable market insights to identify opportunities in existing and potential markets and assess their competitive positions within changing market scenarios. Contact Technavio ResearchJesse MaidaMedia & Marketing ExecutiveUS: +1 844 364 1100UK: +44 203 893 3200Email: [email protected]Website: www.technavio.com/ SOURCE Technavio Report with the AI impact on market trends - The global fast casual restaurants market size is estimated to grow by USD 302.5 billion from 2024-2028, ... Report on how AI is driving market transformation - The global fast fashion market size is estimated to grow by USD 79.2 billion from 2025-2029,... Computer & Electronics Machinery Do not sell or share my personal information:

The red eye that refuses to be extinguished, the metal body that cannot be crushed -- for many of us the word "robot" conjures one image: the Terminator. But robots are now everywhere, serving as companions in care homes or vacuum cleaners in our homes, and manufacturers are keener than ever to design friendly machines. "At first we noticed the kids could be a bit afraid," said Do Hwan Kim of his firm Neubility's tiny delivery robot. To get around the problem, the firm added big doughy eyes that can indicate, making it look like the world's friendliest futuristic wheelie bin. Dozens of the machines now trundle around campsites, university campuses and golf courses across South Korea. "Campsites even use it on their posters," Kim told AFP at the VivaTech trade fair in Paris, underlining its transformation from potential threat to family friend. And VivaTech played host to plenty of other robots designed with cuteness in mind -- ones with cartoon animal personas, others that looked like children's toys from the 1980s. The aesthetic stands in stark contrast to the creepy dog-bots and anonymous drones that have become standard. As robots have become more common, a whole field of academia has grown up studying the interaction between machines and humans. Kerstin Dautenhahn of Waterloo University in Canada, one of the most noted researchers in the area, said she had seen a huge shift in the way manufacturers looked at design: from an all-consuming concern with function to an acute awareness of appearance. "What you find in many, many fields... is that people pay a lot of attention to how the robot moves, how it looks, how it can interact with them," she said. This holds true for robots sharing production lines with human workers just as much as those designed to care for older and disabled people. "Even with those robots where the main function for example is to transport objects from A to B, people still need to pay a lot of attention to how the robot moves, how it can express its intentions," she said. A French firm called Enchanted Tools has fully committed to the friendly aesthetic. Their robots have names, genders, cartoon-style personas and even a back story. "These two characters have escaped from a cartoon to come into our everyday lives to help us manage our social spaces," said the firm's boss Jerome Monceaux. He envisages the brightly coloured machines with cat-like features will help in hospitals, hotels, restaurants and anywhere with objects that need moving. These cute robots take their design cues from a family of social companion robots, which are big business in Japan. Dautenhahn says there is plenty of evidence that people in Japan and South Korea hold more positive views about robots than people in the West. "In Japan, if you say 'I want to build a robot that helps older people in a care home to be happier', they just think it's a great idea," she said. In European countries, the initial response is often negative, fuelled by dystopian science fiction. "We have to do a lot of convincing," she added. Small pilot schemes in the United States have seen robots get bullied or even assaulted, though social-media videos have also shown people helping robots navigate pedestrian crossings. Handling these cultural difficulties is a challenge, says Dautenhahn. But there are plenty of other difficulties. Robots are expensive to design and manufacture, and so they don't come cheap to buy. Enchanted Tools reckons its robots will retail at 35,000 euros ($38,000) while Neubility said it aimed to manufacture its bot for $5,000. Then there is the issue of finding a market. Do Hwan Kim said Neubility was aiming to corner grocery deliveries and has a pilot scheme with the 7-Eleven chain in South Korea. But its robot faces a common hurdle for machines: it cannot climb stairs. Kim hopes market forces will give a helping hand. "At the moment, the delivery cost is so much cheaper with the robot that people are happy to come down the stairs to get their groceries," he said. © Copyright AFP 2025. All rights reserved.

Social companion robots are no longer just science fiction. In classrooms, libraries and homes, these small machines are designed to read stories, play games or offer comfort to children. They promise to support learning and companionship, yet their role in family life often extends beyond their original purpose. In our recent study of families in Canada and the United States, we found that even after a children's reading robot "retired" or was no longer in active and regular use, most households chose to keep it - treating it less like a gadget and more like a member of the family. Luka is a small, owl-shaped reading robot, designed to scan and read picture books aloud, making story time more engaging for young children. In 2021, my colleague Rhonda McEwen and I set out to explore how 20 families used Luka. We wanted to study not just how families used Luka initially, but how that relationship was built and maintained over time, and what Luka came to mean in the household. Our earlier work laid the foundation for this by showing how families used Luka in daily life and how he bond grew over the first months of use. When we returned in 2025 to follow up with 19 of those families, we were surprised by what we found. Eighteen households had chosen to keep Luka, even though its reading function was no longer useful to their now-older children. The robot lingered not because it worked better than before, but because it had become meaningful. Children often spoke about Luka in affectionate, human-like terms. One called it "my little brother." Another described it as their "only pet." These weren't just throwaway remarks -- they reflected the deep emotional place the robot had taken in their everyday lives. Because Luka had been present during important family rituals like bedtime reading, children remembered it as a companion. Parents shared similar feelings. Several explained that Luka felt like "part of our history." For them, the robot had become a symbol of their children's early years, something they could not imagine discarding. One family even held a small "retirement ceremony" before passing Luka on to a younger cousin, acknowledging its role in their household. Other families found new, practical uses. Luka was repurposed as a music player, a night light or a display item on a bookshelf next to other keepsakes. Parents admitted they continued to charge it because it felt like "taking care of" the robot. The device had long outlived its original purpose, yet families found ways to integrate it into daily routines. 'Domesticating' technologies The way participants treated Luka challenges how we usually think about technology, which is that gadgets are disposable. A new phone replaces an old one, toys break and get thrown away and laptops end up in e-waste bins. But when technologies enter family life, especially around emotionally significant moments like story time, they can become part of the household in lasting ways. Our research findings also have important implications for design. Should robots come with an end-of-life plan that recognizes their emotional value? Should companies design with the expectation that some products will be cherished and repurposed, not just discarded and replaced? There are environmental dimensions, too. If families hold on to robots because of attachment, fewer may end up in landfills; this complicates how we think about sustainability and recycling when devices are treated more like keepsakes than tools that may outlive their usefulness. Scholars who study human-computer interaction often use the term "domestication" to describe how technologies become embedded in everyday routines and meanings. More than machines Our study extends that idea to what happens when technology retires. Luka was no longer useful in the conventional sense, but families still made space for it emotionally, symbolically and practically. Many of us keep objects for sentimental reasons, long after they have served their original purpose. Luka shows us that robots can become more than machines. Technology is often framed as fast-moving and disposable. But sometimes, as these families revealed, it lingers. A retired robot can stay in the household because it matters. Zhao Zhao is an assistant professor of computer Science at the University of Guelph. This article is republished from The Conversation under a Creative Commons license. Read the original article. The views and opinions in this commentary are solely those of the writer.

Select Page Thinking about a relationship with a social robot? Markets have a few surprises left…. The man and car of the year – not much of a surprise. PLUS we are now on Spotify and Amazon Music/Podcasts! Click HERE for Show Notes and Links DHUnplugged is now streaming live – with listener chat. Click on link on the right sidebar. Love the Show? Then how about a Donation? Follow John C. Dvorak on Twitter Follow Andrew Horowitz on Twitter Warm Up – Horse can smell the barn – All eyes and $ on Apple – Sloppy trade into the end of the year – Announcing Time Person of The Year – Motor Trend’s Truck of the Year – Inflation – can we talk – One chart – that will blow your mind… Market Update – HUGE MOVE – S&P 500 up 3.8% for the week – BUT, Covid in the headlines again – confusing info about Pfizer efficacy in Israel. BOOSTER – BBB plan – Manchin still unhappy, 1.5% share buyback tax – inflationary or deflationary – Crypto found some legs, but not so supportive – Crytpo segment coming up – HUGE – Fed Rate Decision this week (markets worried) Learned the difference between prosciutto and speck today. – Big Sunday Sauce night at Casita Horowitz this weekend. PSA – Webinar tomorrow at 5PM (Dec 15th) – Register at www.thedisciplinedinvestor.com No Agenda – some really great content and help with all your milk needs – this week – Water Buffalo milk (Thursdays and Sundays) Confusing – S&P trading towards low of day amid concerns about Omicron spreading in the UK (Monday morning) – Last week we were all so excited that there nothing to worry about – Now, vaccine efficacy is reduced – JCD – explain Stop the Press! – The November Producer Price Index showed that the index for final demand increased 0.8% month-over-month while the index for final demand, less foods and energy, increased 0.7% . – That left the year-over-year increases on an unadjusted basis at 9.6% and 7.7%, respectively. PPI Final 1 PPI Final 2 Lies, Damned Lies and The Fed – Only one of two things can be true – 1) The Fed was lying about their inflation outlook to keep markets happy (manipulation) – 2) The Fed was truthful and that confirms that they have no ability to forecast (stupidity and can’t be trusted) – – – Either way, they can’t be trusted.. (although they do hold the purse strings) Robinhood Stock – $HOOD – Looks like robbed from the poor and gave to the rich with this failed IPO/public entry. HOOD Chart ELON – Musk is “thinking of” leaving his jobs and becoming an influencer – “It would be nice to have a bit more free time on my hands as opposed to just working day and night, from when I wake up to when I go to sleep 7 days a week. Pretty intense.” – Also – Just announced – Time magazine’s person of the Year In an Odd Twist to the Above – MotorTrend on Monday named the all-electric Rivian R1T the publication’s 2022 truck of the year, beating out other pickups from Ford Motor, General Motors and Hyundai Motor. – MotorTrend called the R1T, which is the first mass-produced electric truck in the U.S., “the most remarkable pickup truck we’ve ever driven,” in a release announcing the award. -MotorTrend said the Rivian R1T excelled in each of its six key criteria: safety, efficiency, value, advancement in design, engineering excellence, and performance of intended function. – Other finalists for MotorTrend’s truck of the year were the Ford Maverick, GMC Hummer EV and Hyundai Santa Cruz. Say What? – Fox News anchor Chris Wallace abruptly announced on Sunday he’s leaving the network after 18 years, effective immediately. – The host of the flagship “Fox News Sunday” said he was ready for a change. – “I want to try something new, to go beyond politics to all the things I’m interested in. I’m ready for a new adventure,” Wallace said in a statement that aired on his final show. Wallace didn’t provide additional details on his new endeavor, but said he hoped fans would “check it out.” – GOING TO CNN Turkish Lira – ATL – The Turkish lira crashed as much as 7% in just a few minutes to a new record near 15 to the dollar on Monday, gripped by worries over President Tayyip Erdogan’s risky new economic policy and prospects of another interest rate cut on Thursday. – Concern that Erdogan will lower rates by 1% in the face of 20% inflation – Turkish Stock Market ETF – TUR Home/Social Robots – JCD?? You Getting one? – Amazon’s Astro – Seems more like a moving security camera – RING capabilities (Camera, Remote, Alerts) – Social robots are designed to engage with people more as a collaborative partner,” said Cynthia Breazeal, director of the Personal Robots Group at the MIT Media Lab. “As opposed to a tool that you use, you interact more in an interpersonal way to achieve tasks or goals or experiences.” – SOCIAL ROBOTS – Think Sleeper – Woody Allen – “People want to have longer, more meaningful, more interesting conversations with these technologies. They get frustrated when it’s too transactional,” Breazeal said. “I think there’s a hunger and a desire for people to be able to interact with these technologies in this way.” One Chart…. – The average Nasdaq Composite stock is 39% below its 52-week high even though the index is just 3.6% below its 52-week high. Average Stock Distance MetaVerse Alert – Snoop Dogg is developing a Snoopverse – An NFT collector spent a little under a half-million dollars for the privilege of becoming Snoop Dogg’s next-door neighbor – In the metaverse – Snoop Dog building a virtual world in the Sandbox – “I’m always on the lookout for new ways of connecting with fans and what we’ve created in The Sandbox is the future of virtual hangouts, NFT drops, and exclusive concerts,” Snoop Dogg said in a press release, according to Decrypt. – MORE: Facebook on Thursday announced that it is opening up Horizon Worlds, its virtual reality world of avatars, to anyone 18 and older in the U.S. and Canada. —– In Horizon Worlds, users of Facebook’s Oculus virtual reality headsets can create a legless avatar to wander in the animated virtual world. There, they can play games and interact with other users’ avatars. METAVERSE – Since Facebook announced its switch to Meta and its future plans to create a metaverse, the existing ones have gained popularity. The biggest racked up $ 100 million selling NFTs the week of November 22 and November 28. – The activity continues with these platforms such as The Sandbox, Decenterland, CryptoVoxels and Somnium Space. – SANDBOX biggest player right now More Metaverse – Metaverse infrastructure platform for brands, InfiniteWorld has announced that it is going public via a Special Purpose Acquisition Company (SPAC), Aries, at a valuation of $700 million and will be listed on the Nasdaq under the ticker “JPG.” – InfiniteWorld provides the infrastructure to create digital assets and NFTs and engage with customers in the Metaverse. Its NFT architecture is built on top of the SUKU protocol and provides a secure transfer of ownership of digital assets. And this… – Nike just bought a virtual shoe company that makes NFTs and sneakers ‘for the metaverse’ – Hong Kong property tycoons, brokers snap up virtual land in metaverse as valuations soar – Virtual land sales have become one of the hottest new trends in a tech world increasingly fascinated with the metaverse, a shared, immersive 3D space where people can interact. Valuations for virtual land at times exceed those in the real world. Doomsday Alert! – Robert Kiyosaki, the author of Rich Dad Poor Dad, has once again warned of an impending crash, followed by a depression. He expects several markets, including bitcoin, to crash. – Tweet: Crash and depression coming. Gold, silver, bitcoin, real estate will crash too. Ready to buy more gold, silver, bitcoin, real estate after crash has crashed. Time to get richer after fake inflation crashes. More Predictions from Kiyosaki – “Biden to step down soon,” he predicted. “Kamala [Harris] will become [the] first female president. This was [the] plan all along. Trump ruined Hillary’s ascension following Obama.” Retail Changes Coming – Massive space available at malls – no news here – 90 million sq feet available (16 Mall of the Americas) – Looking to now rent to casinos, amusement parks, medical facilities, storage units, hotels, schools, offices and residences Crypto Segment – Dave Ackerman – After being sluggish for most of the week, Bitcoin price jumped briefly on news of U.S. inflation rate hitting a 40 year high. The relative strength index (RSI) on the daily chart last week was the most oversold since September, which was followed at that time by a rally in price. – Bitcoin hash rate returned to all-time highs recovering from a low in June after a crackdown on mining in China. Many believe that hash rate corresponds with price action, suggesting a possible near term increase in price despite gloomy sentiment. – FreeRossDAO emerged as the winner in the auction for imprisoned Silk Road founder Ross Ulbricht’s collection of NFT’s with a bid of 1,446 Ethereum tokens. The organization stated that it raised over 2,800 Eth ($12 million) from more than 1,300 supporters through the PleasrDAO community —- Thoughts – DAOs -Pricing Consortium – potential high for manipulation even though it seems like a community thing. – Decentralized autonomous organizations (is it really autonomous and decentralized?) It looks like a mob with money. Or investment club. – Tesla announced it will accept Dogecoin for payment of Tesla merchandise. Thinking about a relationship with a social robot? Markets have a few surprises left…. The man and car of the year – not much of a surprise. PLUS we are now on Spotify and Amazon Music/Podcasts! DHUnplugged is now streaming live – with listener chat. Click on link on the right sidebar. Then how about a Donation? Follow John C. Dvorak on Twitter Follow Andrew Horowitz on Twitter Warm Up – Horse can smell the barn – All eyes and $ on Apple – Sloppy trade into the end of the year – Announcing Time Person of The Year – Motor Trend’s Truck of the Year – Inflation – can we talk – One chart – that will blow your mind… Market Update – HUGE MOVE – S&P 500 up 3.8% for the week – BUT, Covid in the headlines again – confusing info about Pfizer efficacy in Israel. BOOSTER – BBB plan – Manchin still unhappy, 1.5% share buyback tax – inflationary or deflationary – Crypto found some legs, but not so supportive – Crytpo segment coming up – HUGE – Fed Rate Decision this week (markets worried) Learned the difference between prosciutto and speck today. – Big Sunday Sauce night at Casita Horowitz this weekend. PSA – Webinar tomorrow at 5PM (Dec 15th) – Register at www.thedisciplinedinvestor.com No Agenda – some really great content and help with all your milk needs – this week – Water Buffalo milk (Thursdays and Sundays) Confusing – S&P trading towards low of day amid concerns about Omicron spreading in the UK (Monday morning) – Last week we were all so excited that there nothing to worry about – Now, vaccine efficacy is reduced – JCD – explain Stop the Press! – The November Producer Price Index showed that the index for final demand increased 0.8% month-over-month while the index for final demand, less foods and energy, increased 0.7% . – That left the year-over-year increases on an unadjusted basis at 9.6% and 7.7%, respectively. PPI Final 1 PPI Final 2 Lies, Damned Lies and The Fed – Only one of two things can be true – 1) The Fed was lying about their inflation outlook to keep markets happy (manipulation) – 2) The Fed was truthful and that confirms that they have no ability to forecast (stupidity and can’t be trusted) – – – Either way, they can’t be trusted.. (although they do hold the purse strings) Robinhood Stock – $HOOD – Looks like robbed from the poor and gave to the rich with this failed IPO/public entry. HOOD Chart ELON – Musk is “thinking of” leaving his jobs and becoming an influencer – “It would be nice to have a bit more free time on my hands as opposed to just working day and night, from when I wake up to when I go to sleep 7 days a week. Pretty intense.” – Also – Just announced – Time magazine’s person of the Year In an Odd Twist to the Above – MotorTrend on Monday named the all-electric Rivian R1T the publication’s 2022 truck of the year, beating out other pickups from Ford Motor, General Motors and Hyundai Motor. – MotorTrend called the R1T, which is the first mass-produced electric truck in the U.S., “the most remarkable pickup truck we’ve ever driven,” in a release announcing the award. -MotorTrend said the Rivian R1T excelled in each of its six key criteria: safety, efficiency, value, advancement in design, engineering excellence, and performance of intended function. – Other finalists for MotorTrend’s truck of the year were the Ford Maverick, GMC Hummer EV and Hyundai Santa Cruz. Say What? – Fox News anchor Chris Wallace abruptly announced on Sunday he’s leaving the network after 18 years, effective immediately. – The host of the flagship “Fox News Sunday” said he was ready for a change. – “I want to try something new, to go beyond politics to all the things I’m interested in. I’m ready for a new adventure,” Wallace said in a statement that aired on his final show. Wallace didn’t provide additional details on his new endeavor, but said he hoped fans would “check it out.” – GOING TO CNN Turkish Lira – ATL – The Turkish lira crashed as much as 7% in just a few minutes to a new record near 15 to the dollar on Monday, gripped by worries over President Tayyip Erdogan’s risky new economic policy and prospects of another interest rate cut on Thursday. – Concern that Erdogan will lower rates by 1% in the face of 20% inflation – Turkish Stock Market ETF – TUR Home/Social Robots – JCD?? You Getting one? – Amazon’s Astro – Seems more like a moving security camera – RING capabilities (Camera, Remote, Alerts) – Social robots are designed to engage with people more as a collaborative partner,” said Cynthia Breazeal, director of the Personal Robots Group at the MIT Media Lab. “As opposed to a tool that you use, you interact more in an interpersonal way to achieve tasks or goals or experiences.” – SOCIAL ROBOTS – Think Sleeper – Woody Allen – “People want to have longer, more meaningful, more interesting conversations with these technologies. They get frustrated when it’s too transactional,” Breazeal said. “I think there’s a hunger and a desire for people to be able to interact with these technologies in this way.” One Chart…. – The average Nasdaq Composite stock is 39% below its 52-week high even though the index is just 3.6% below its 52-week high. Average Stock Distance MetaVerse Alert – Snoop Dogg is developing a Snoopverse – An NFT collector spent a little under a half-million dollars for the privilege of becoming Snoop Dogg’s next-door neighbor – In the metaverse – Snoop Dog building a virtual world in the Sandbox – “I’m always on the lookout for new ways of connecting with fans and what we’ve created in The Sandbox is the future of virtual hangouts, NFT drops, and exclusive concerts,” Snoop Dogg said in a press release, according to Decrypt. – MORE: Facebook on Thursday announced that it is opening up Horizon Worlds, its virtual reality world of avatars, to anyone 18 and older in the U.S. and Canada. —– In Horizon Worlds, users of Facebook’s Oculus virtual reality headsets can create a legless avatar to wander in the animated virtual world. There, they can play games and interact with other users’ avatars. METAVERSE – Since Facebook announced its switch to Meta and its future plans to create a metaverse, the existing ones have gained popularity. The biggest racked up $ 100 million selling NFTs the week of November 22 and November 28. – The activity continues with these platforms such as The Sandbox, Decenterland, CryptoVoxels and Somnium Space. – SANDBOX biggest player right now More Metaverse – Metaverse infrastructure platform for brands, InfiniteWorld has announced that it is going public via a Special Purpose Acquisition Company (SPAC), Aries, at a valuation of $700 million and will be listed on the Nasdaq under the ticker “JPG.” – InfiniteWorld provides the infrastructure to create digital assets and NFTs and engage with customers in the Metaverse. Its NFT architecture is built on top of the SUKU protocol and provides a secure transfer of ownership of digital assets. And this… – Nike just bought a virtual shoe company that makes NFTs and sneakers ‘for the metaverse’ – Hong Kong property tycoons, brokers snap up virtual land in metaverse as valuations soar – Virtual land sales have become one of the hottest new trends in a tech world increasingly fascinated with the metaverse, a shared, immersive 3D space where people can interact. Valuations for virtual land at times exceed those in the real world. Doomsday Alert! – Robert Kiyosaki, the author of Rich Dad Poor Dad, has once again warned of an impending crash, followed by a depression. He expects several markets, including bitcoin, to crash. – Tweet: Crash and depression coming. Gold, silver, bitcoin, real estate will crash too. Ready to buy more gold, silver, bitcoin, real estate after crash has crashed. Time to get richer after fake inflation crashes. More Predictions from Kiyosaki – “Biden to step down soon,” he predicted. “Kamala [Harris] will become [the] first female president. This was [the] plan all along. Trump ruined Hillary’s ascension following Obama.” Retail Changes Coming – Massive space available at malls – no news here – 90 million sq feet available (16 Mall of the Americas) – Looking to now rent to casinos, amusement parks, medical facilities, storage units, hotels, schools, offices and residences Crypto Segment – Dave Ackerman – After being sluggish for most of the week, Bitcoin price jumped briefly on news of U.S. inflation rate hitting a 40 year high. The relative strength index (RSI) on the daily chart last week was the most oversold since September, which was followed at that time by a rally in price. – Bitcoin hash rate returned to all-time highs recovering from a low in June after a crackdown on mining in China. Many believe that hash rate corresponds with price action, suggesting a possible near term increase in price despite gloomy sentiment. – FreeRossDAO emerged as the winner in the auction for imprisoned Silk Road founder Ross Ulbricht’s collection of NFT’s with a bid of 1,446 Ethereum tokens. The organization stated that it raised over 2,800 Eth ($12 million) from more than 1,300 supporters through the PleasrDAO community —- Thoughts – DAOs -Pricing Consortium – potential high for manipulation even though it seems like a community thing. – Decentralized autonomous organizations (is it really autonomous and decentralized?) It looks like a mob with money. Or investment club. – Tesla announced it will accept Dogecoin for payment of Tesla merchandise. 420 Stocks … Stocks to watch: CGC, STZ, TLRY, CRON, MJ, NBEV *** NEW INTERACTIVE CHARTS *** CLICK HERE FOR MORE CHARTS ON TRADING VIEW UPDATE – NEW ETF – Global X Millennials Thematic ETF Challenge! The top holdings of Global X Millennials Thematic ETF (MILN) include stocks many investors would expect, such as Facebook, LinkedIn and Amazon, which take advantage of the tech tendencies of millennials. But some names might be more surprising like real estate investment trusts AvalonBay Communities and Equity Residential, and Home Depot, which could benefit from millennials moving out of the home of their parents. We are creating the DH Old Codger Index Portfolio to compete against this new ETF to see how “old school” stocks do in comparison. Companies in our index will include: (updated names as of 8/29/16) We have the performance summary running (daily and since inception of 5/6/16) – DHOCI vs. Millennials ETF Battle JCD Score () See this week’s stock picks HERE Follow John C. Dvorak on Twitter Follow Andrew Horowitz on Twitter Share: (Additional Disclosures and Terms of Use) Sign up to be the first to receive the latest analysis and podcast updates ! Copyright © 2026 The Disciplined Investor | A Publication of Horowitz & Company, Inc. We will email you when a new podcast is released and other items of interest…

This doctoral thesis addresses the challenge of building emotionally intelligent and multilingual conversational systems capable of operating in real-time and physically embodied scenarios. As social robots become more present in healthcare, education, and assistive contexts, there is a growing demand for conversational agents that can understand and respond to human emotions across different languages and cultural settings. Traditional dialogue systems often fall short in managing emotional complexity, maintaining engagement, and adapting to multilingual environments. This research proposes a unified framework that integrates Large Language Models, Reinforcement Learning, and Fuzzy Logic to support emotional, multilingual, and embodied Human-Robot Interaction. The objectives of the thesis are threefold. First, to investigate the design and evaluation of multilingual dialogue resources with emotional annotation, ensuring diversity, realism, and consistency. Second, to explore and develop a dialogue management system for multilingual interaction, integrating contextual and emotional signals into modular response generation. This includes the research and development of an emotion-aware conversational agent to assess emotional alignment, generate empathetic engagement, and produce contextually appropriate responses. Third, to design, implement, and evaluate the full system in social robots, using interpretable emotional reasoning based on Fuzzy Logic and multimodal inputs such as speech, touch, light, and physiological data. The findings are presented in three main contributions. First, a novel method for generating emotional dialogue datasets in English and Spanish is introduced, using Chain-of-Emotion prompting and AI-human preference alignment for training robust models. Second, an emotionally sensitive and multilingual dialogue architecture is implemented, combining Supervised Fine-Tuning, optimization-based Reinforcement Learning, and hierarchical topic and emotion tracking. Third, a Fuzzy Logic Systems based emotional model is extended and integrated into physical robots, supporting real-time emotional reasoning and expressive behavior through structured stimuli-state-expression mappings. Results from the dialogue management system demonstrate that the hierarchical architecture effectively integrates emotional and contextual information to generate coherent and affectively aligned responses across languages. The system was deployed on two robotic platforms and evaluated through simulation and real-world interactions. Results show that the proposed models achieve emotionally aligned responses, support bilingual dialogue, and exhibit consistent internal emotional states that influence expressive output. User studies confirm improved engagement and affective perception. This thesis contributes a modular and interpretable framework for emotionally intelligent and multilingual conversational agents. The proposed emotional model incorporates 43 fuzzy rule tables and 17 fuzzy variables across multiple emotional state dimensions. In addition, an emotionally aligned dialogue dataset with AI feedback was created, containing 128,125 winner-loser preference pairs, to train emotional models using Reinforcement Learning for generating emotionally engaging responses. The system was also evaluated through a user study involving 66 human participants. In particular, participants correctly recognized the robot's emotional expression with accuracy rates of up to 72.7%, with consistent performance across neutral, positive, and negative conditions. By integrating synthetic emotional data generation, emotionally aware model training, and embodied emotional reasoning, the system advances the development of scalable and human-aligned social robots suitable for real-world deployment in sensitive domains. RESUMEN Esta tesis doctoral aborda el desafío de construir sistemas conversacionales emocionalmente inteligentes y multilingües, capaces de operar en tiempo real y en escenarios físicamente embebidos. A medida que los robots sociales se integran en contextos de atención sanitaria, educación y asistencia, crece la demanda de agentes conversacionales que comprendan y respondan a emociones humanas en distintos idiomas y culturas. Los sistemas de diálogo tradicionales suelen fallar al manejar la complejidad emocional, mantener la implicación del usuario y adaptarse a entornos multilingües. Esta investigación propone un marco unificado que combina Modelos Extensos de Lenguaje, Aprendizaje por Refuerzo y Lógica Difusa para fomentar una Interacción Humano-Robot emocional, multilingüe y contextualizada. Los objetivos de la tesis son tres. Primero, investigar el diseño y evaluación de recursos de diálogo multilingües con anotaciones emocionales, garantizando diversidad, realismo y coherencia. Segundo, desarrollar un sistema de gestión del diálogo que integre señales contextuales y afectivas en una generación modular de respuestas. Esto incluye un agente conversacional consciente de las emociones, capaz de evaluar la alineación emocional, generar implicación empática y producir respuestas apropiadas al contexto. Tercero, implementar y evaluar el sistema completo en robots sociales, mediante razonamiento emocional interpretable basado en Lógica Difusa y entradas multimodales como voz, táctil, luz y señales fisiológicas. Los resultados se organizan en tres contribuciones. Primero, se introduce un método para generar datos de diálogo emocional en inglés y español, utilizando el esquema Cadena-de-Emociones y un procedimiento de alineación de preferencias entre IA y humanos. Segundo, se implementa una arquitectura de diálogo multilingüe y emocionalmente sensible, que combina Ajuste Fino Supervisado, Aprendizaje por Refuerzo optimizado y clasificación jerárquica de temas y emociones. Tercero, se complementa un modelo emocional basado en Sistemas de Lógica Difusa, que permite razonamiento afectivo en tiempo real y expresión emocional mediante reglas estructuradas entre estímulo, estado y expresión. Los resultados del sistema de gestión del diálogo demuestran que la arquitectura jerárquica integra eficazmente información emocional y contextual para generar respuestas coherentes y alineadas en varios idiomas. El sistema fue desplegado en dos robots y evaluado mediante simulaciones e interacciones reales. Los modelos generaron respuestas emocionalmente alineadas, permitieron diálogo multilingüe y mantuvieron estados emocionales internos consistentes. Estudios con usuarios confirmaron una mayor implicación y mejor percepción afectiva. La tesis aporta un marco modular e interpretable para agentes conversacionales emocionalmente inteligentes y multilingües. El modelo emocional incluye 43 tablas de reglas difusas y 17 variables difusas en distintas dimensiones emocionales. Además, se creó un conjunto de datos alineado emocionalmente mediante retroalimentación de IA, con 128.125 pares de preferencia ganadorperdedor, empleado para entrenar modelos mediante Aprendizaje por Refuerzo que generan respuestas emocionales y atractivas. El sistema fue evaluado con 66 participantes humanos, quienes reconocieron correctamente la expresión emocional del robot con una precisión de hasta el 72,7%, con un rendimiento constante en las condiciones neutral, positiva y negativa. Mediante la integración de la generación sintética de datos emocionales, el entrenamiento de modelos conscientes de las emociones y el razonamiento emocional personificado, el sistema impulsa el desarrollo de robots sociales escalables y alineados con el ser humano, adecuados para su despliegue en entornos sensibles del mundo real. El Archivo Digital UPM es el repositorio digital institucional mantenido por la Biblioteca de la Universidad Politécnica de Madrid. Desarrollado y gestionado con EPrints. Sindicación: Atom, RSS 2.0 y RSS 1.0 (HTML) Recolección: OAI 2.0 El Archivo Digital UPM es el repositorio digital institucional mantenido por la Biblioteca de la Universidad Politécnica de Madrid. Desarrollado y gestionado con EPrints. Sindicación: Atom, RSS 2.0 y RSS 1.0 (HTML) Recolección: OAI 2.0

This article by Shane Saunderson, University of Toronto originally appeared on the Conversation and is published here with permission. In the mid-1990s, there was research going on at Stanford University that would change the way we think about computers. The Media Equation experiments were simple: participants were asked to interact with a computer that acted socially for a few minutes after which, they were asked to give feedback about the interaction. Participants would provide this feedback either on the same computer (No. 1) they had just been working on or on another computer (No. 2) across the room. The study found that participants responding on computer No. 2 were far more critical of computer No. 1 than those responding on the same machine they’d worked on. People responding on the first computer seemed to not want to hurt the computer’s feelings to its face, but had no problem talking about it behind its back. This phenomenon became known as the computers as social actors (CASA) paradigm because it showed that people are hardwired to respond socially to technology that presents itself as even vaguely social. The CASA phenomenon continues to be explored, particularly as our technologies have become more social. As a researcher, lecturer and all-around lover of robotics, I observe this phenomenon in my work every time someone thanks a robot, assigns it a gender or tries to justify its behaviour using human, or anthropomorphic, rationales. What I’ve witnessed during my research is that while few are under any delusions that robots are people, we tend to defer to them just like we would another person. Social tendencies While this may sound like the beginnings of a Black Mirror episode, this tendency is precisely what allows us to enjoy social interactions with robots and place them in caregiver, collaborator or companion roles. The positive aspects of treating a robot like a person is precisely why roboticists design them as such — we like interacting with people. As these technologies become more human-like, they become more capable of influencing us. However, if we continue to follow the current path of robot and AI deployment, these technologies could emerge as far more dystopian than utopian. The Sophia robot, manufactured by Hanson Robotics, has been on 60 Minutes, received honorary citizenship from Saudi Arabia, holds a title from the United Nations and has gone on a date with actor Will Smith. While Sophia undoubtedly highlights many technological advancements, few surpass Hanson’s achievements in marketing. If Sophia truly were a person, we would acknowledge its role as an influencer. However, worse than robots or AI being sociopathic agents — goal-oriented without morality or human judgment — these technologies become tools of mass influence for whichever organization or individual controls them. If you thought the Cambridge Analytica scandal was bad, imagine what Facebook’s algorithms of influence could do if they had an accompanying, human-like face. Or a thousand faces. Or a million. The true value of a persuasive technology is not in its cold, calculated efficiency, but its scale. Seeing through intent Recent scandals and exposures in the tech world have left many of us feeling helpless against these corporate giants. Fortunately, many of these issues can be solved through transparency. There are fundamental questions that are important for social technologies to answer because we would expect the same answers when interacting with another person, albeit often implicitly. Who owns or sets the mandate of this technology? What are its objectives? What approaches can it use? What data can it access? Since robots could have the potential to soon leverage superhuman capabilities, enacting the will of an unseen owner, and without showing verbal or non-verbal cues that shed light on their intent, we must demand that these types of questions be answered explicitly. As a roboticist, I get asked the question, “When will robots take over the world?” so often that I’ve developed a stock answer: “As soon as I tell them to.” However, my joke is underpinned by an important lesson: don’t scapegoat machines for decisions made by humans. I consider myself a robot sympathizer because I think robots get unfairly blamed for many human decisions and errors. It is important that we periodically remind ourselves that a robot is not your friend, your enemy or anything in between. A robot is a tool, wielded by a person (however far removed), and increasingly used to influence us. Shane Saunderson, Ph.D. Candidate, Robotics, University of Toronto This article is republished from The Conversation under a Creative Commons license. Read the original article. If you would like to apply to become a Verified Commenter, please fill out this form. More Spotlight > © 2026 OrilliaMatters.com

Searching for your content... In-Language News Contact Us 888-776-0942 from 8 AM - 10 PM ET Oct 27, 2021, 21:00 ET Share this article NEW YORK, Oct. 27, 2021 /PRNewswire/ -- The "Social Robots Market by Component (Hardware, Software, and Services) and Geography (APAC, North America, Europe, MEA, and South America) - Forecast and Analysis 2021-2025" report has been added to Technavio's offerings. With ISO 9001:2015 certification, Technavio has been proudly partnering with more than 100 Fortune 500 companies for over 16 years. The potential growth difference for the social robots market between 2020 and 2025 is USD 1.10 bn. To get the exact yearly growth variance and the Y-O-Y growth rate, Talk to our analyst. Key Market Dynamics: The technological advances in social robots and increasing government support for the development of advanced social robots are some of the key market drivers. However, factors such as high cost will challenge market growth. To learn about additional key drivers, trends, and challenges available with Technavio, Read our FREE Sample Report right now! The social robots market report is segmented by component (hardware, software, and services) and geography (APAC, North America, Europe, MEA, and South America). APAC will be the leading region with 36% of the market's growth during the forecast period. China, Japan, and South Korea (Republic of Korea) are the key markets for social robots in APAC. View our sample report for additional insights into the contribution of all the segments and regional opportunities in the report. Some Companies Mentioned Related Reports: Social Robots Market Scope Report Coverage Details Page number 120 Base year 2020 Forecast period 2021-2025 Growth momentum & CAGR Accelerate at a CAGR of 14.43% Market growth 2021-2025 USD 1.10 billion Market structure Fragmented YoY growth (%) 13.10 Regional analysis APAC, North America, Europe, MEA, and South America Performing market contribution APAC at 36% Key consumer countries US, China, UK, Japan, and South Korea (Republic of Korea) Competitive landscape Leading companies, competitive strategies, consumer engagement scope Companies profiled BLUE FROG ROBOTICS SAS, Furhat Robotics AB, Hitachi Ltd., Knightscope Inc., Nippon Telegraph and Telephone Corp., PAL Robotics SL, Savioke Inc., SoftBank Group Corp., and Ubtech Robotics Inc. Market Dynamics Parent market analysis, market growth inducers and obstacles, fast-growing and slow-growing segment analysis, COVID-19 impact and future consumer dynamics, market condition analysis for the forecast period. Customization purview If our report has not included the data that you are looking for, you can reach out to our analysts and get segments customized. Key Topics Covered: About UsTechnavio is a leading global technology research and advisory company. Their research and analysis focus on emerging market trends and provide actionable insights to help businesses identify market opportunities and develop effective strategies to optimize their market positions. With over 500 specialized analysts, Technavio's report library Their client base consists of enterprises of all sizes, including more than 100 Fortune 500 companies. This growing client base relies on Technavio's comprehensive coverage, extensive research, and actionable market insights to identify opportunities in existing and potential markets and assess their competitive positions within changing market scenarios. ContactTechnavio ResearchJesse MaidaMedia & Marketing ExecutiveUS: +1 844 364 1100UK: +44 203 893 3200Email:[email protected]Website: www.technavio.com/ SOURCE Technavio Report with the AI impact on market trends - The global fast casual restaurants market size is estimated to grow by USD 302.5 billion from 2024-2028, ... Report on how AI is driving market transformation - The global fast fashion market size is estimated to grow by USD 79.2 billion from 2025-2029,... Machinery Mining & Metals Mining & Metals Do not sell or share my personal information:

In the evolving world of educational technology, a recent study is turning heads by suggesting that social robots could be key allies in helping children overcome the jitters associated with reading aloud. Researchers at the University of Chicago have delved into how these nonjudgmental machines might provide a safer space for young learners to practice, potentially transforming classroom dynamics. The findings, detailed in a paper published in Science Robotics, indicate that kids exhibit fewer physiological signs of anxiety—like reduced skin conductance—when reading to a robot compared to a human. The study involved elementary school students who read passages to either a small robot companion or an adult listener. By measuring stress indicators and self-reported feelings, the team found that robots fostered a more relaxed environment, encouraging fluency without the fear of criticism. This isn’t just about tech novelty; it’s rooted in human-robot interaction principles, where machines can offer consistent, patient feedback. Exploring the Mechanics of Robot-Assisted Learning At the heart of this research is the Human-Robot Interaction Lab at the University of Chicago, led by Assistant Professor Sarah Sebo. PhD student Lauren Wright, who spearheaded the project, collaborated with experts from the University of Illinois Chicago and the University of Wisconsin–Madison to design experiments that mimic real classroom scenarios. As reported in University of Chicago News, the robots used were programmed to respond empathetically, nodding or making encouraging sounds, which helped build children’s confidence over time. This approach addresses a persistent issue: reading anxiety can linger into adulthood, hindering literacy development. The study’s participants, aged around 8 to 10, showed marked improvements in engagement when interacting with robots, suggesting these devices could supplement teachers by handling repetitive practice sessions. Broader Implications for Educational Tech Integration Industry experts are buzzing about the potential scalability. According to CNET, which highlighted the study, robots might offer “unique support” in learning environments, especially for children with social anxieties or learning disabilities. This aligns with prior research, such as a 2018 piece in Popular Science that explored robots turning solo reading into interactive activities to boost motivation at home. However, challenges remain. Deploying robots in schools requires addressing costs, privacy concerns, and teacher training. The UChicago team emphasizes that robots aren’t replacements for human educators but tools to augment their efforts, particularly in under-resourced districts where individualized attention is scarce. From Lab to Classroom: Real-World Applications and Future Directions Looking ahead, the study opens doors to more sophisticated AI-driven companions. For instance, integrating natural language processing could allow robots to provide real-time pronunciation tips or comprehension questions, as hinted in related work from UChicago’s Department of Computer Science. Early adopters, like pilot programs in libraries, are already testing these ideas, with feedback indicating higher reading enthusiasm among kids. Critics, though, warn of over-reliance on tech, stressing the need for balanced human interaction. Yet, as anxiety affects millions of students—potentially stalling academic progress—these findings could inspire a new wave of edtech investments. Companies developing social robots, from startups to giants like SoftBank, are likely watching closely, eyeing integrations that blend empathy with education. Ethical Considerations and Long-Term Impact Ethically, ensuring equitable access is crucial; not every school can afford high-end robots. The research also touches on emotional bonds, as noted in a TechXplore article about robots gaining “emotional value” in daily rituals. For children, this could mean forming positive associations with learning, reducing dropout risks in literacy programs. Ultimately, this UChicago-led initiative underscores a pivotal shift: technology isn’t just about efficiency but emotional support. As more studies build on these results, we may see robots becoming standard classroom fixtures, helping a generation read with confidence rather than fear. Subscribe for Updates The AITrends Email Newsletter keeps you informed on the latest developments in artificial intelligence. Perfect for business leaders, tech professionals, and AI enthusiasts looking to stay ahead of the curve. Help us improve our content by reporting any issues you find. Get the free daily newsletter read by decision makers Get our media kit Deliver your marketing message directly to decision makers.

Searching for your content... In-Language News Contact Us 888-776-0942 from 8 AM - 10 PM ET Jun 23, 2021, 11:15 ET Share this article DUBLIN, June 23, 2021 /PRNewswire/ -- The "Social Robots Market - Forecasts from 2021 to 2026" report has been added to ResearchAndMarkets.com's offering. The social robots market is expected to grow at a compound annual growth rate of 12.68% over the analyzed period to reach a market size of US$912.488 million in 2026 from US$395.577 million in 2019. Social Robots are companion robots. These robots help in lifting, companionship, and simulation of emotions in human beings. In addition, these robots assist in everyday life, by performing various activities. Sony was the first company that introduced social robots in 1999. AIBO is a social robot pet dog that responds to its owner's voice and action. This robot has significant market standing even 20 years later of its first launch. The software of AIBO was last updated in April 2021, to accommodate new features such as app connection.In 2006, France-based Aldebaran Robotic launched the Nao robot for therapy of autistic children. The first nursing social robot in Japan was launched in 2015. Developed by the Riken institute, Robear assists patients and caregivers in a nursing home across Japan. The robot assists in lifting patients from their beds by providing physical strength. Growing complexities of the world and increased adoption of automation will drive the demand for social robots. In malls and shopping complexes, social robots guide the customers, increasing their shopping experience. In parking lots, robots assist in smooth functioning. The Healthcare sector also employs social robots for better treatment of patients. Furthermore, robots ease household chores and provide companionship to the owner. However, growing automation results in a decrease in employment opportunities and have raised several concerns, hindering the market growth.Social Robots assist in the caregiving of the aging population and facilitate better mental health in specially-abled people hence driving the adoption of robots in the healthcare sector.Based on end-users, the social robot market is segmented into healthcare, education, retail, entertainment, IT and communication, household, and others. In the retail and entertainment sector, social robots are used to assist and guide the customers towards their desired choices. Many giant retail shops, in the US, have installed social robots which help customers find what they need by providing the required information. Mobile stores across Japan have installed more than 140 SoftBank's Pepper Robots since its launch. Success in enhancing the consumer experience, SoftBank received an investment of US$50 billion from Apple, in 2020, for research and development on Pepper. Target, US, installed Tally, a social robot build by Simbe Robotics, for inventory management. In 2016, a leading fast-food chain, Dominos, Australia, announced the introduction of drive-less vehicles, Domino's Robotic Unit (DRU) for better delivery of pizza. Growth in the IT and communication industry will also increase the adoption of social robots to cater to customer queries and needs, which will provide significant growth prospects. However, during the forecasted period, healthcare and household sectors are projected to hold dominating share of the market. In the healthcare sector, social robots are used for nursing purposes. These robots also help in catering emotional needs of the patients. Japan is the leading user of social robots in its healthcare sector. The education sector will also grow at a significant rate.The rise in automation will contribute to the market growth during the forecasted period.The key factor driving the growth in the market is surging automation across varied industrial verticals. The International Federation of Robotics has predicted that the adoption of robots across many industry verticals will grow at a quadruple rate in the coming decade. In 2017, robot installation increased 21% in Asia, 16% in the Americas, and 8% in the European region. China, South Korea, Japan, United States, Germany are the top 5 robotics markets, accounting for 74% of the total robotics supply (2016). A significant proportion is a share by the social robot. Hence, growth in the robotic industry will support the market growth of social robots.The Asia Social Robots market will grow at an exponential rate with Japan dominating the global healthcare social robot market.Based on geography, the social robot market is divided into North America, South America, Europe, the Middle East and Africa, and the Asia Pacific region. The North America and Asia Pacific region are the leading social robot market and will dominate the global market by the end of 2026. Particularly, the Asia region has vast adoption of robots in the healthcare and education sector, with Japan and South Korea being the prime market for healthcare social robots.In Japan, the aging population has been on a rise and is increasing at a quadruple rate. While the number of people aged 65 and above accounted for 19.098% of the population, the proportion had increased to 28.002%. while the population is falling at 1.4% annually. A large aged population has increased the requirement for nursing and other facilities while the workforce is limited. Hence, Japan has increasingly adopted social robots for better care and nursing of aged people. An investment of US$ 100 million was made by the government in 2018 for nursing social robot development and installation. Further, Japan is a leader in the development of healthcare and nursing robots. Paro, Telenoid, and Ugo are leading social nursing robots in Japan. South Korea is another emerging market for healthcare robots. Moreover, the preferability of the aging population for automation over immigration provides a stable market. Japan has also intended to increase the involvement of social robots in its education sector.Estimates by World Bank show that South Korea will surpass Japan in the proportion of the aged population (above 65) to reach a proportion around 37% by 2045. Japan, on the other hand, will have 36.7% of its population in the category in 2045. The growing aged population and rising automation have increased concerns by the government, resulting in the announcement of an investment of US$250 million in automation. in March 2020, with a prime focus on healthcare, disaster response, and rehabilitation.COVID-19 InsightsThe effect of the coronavirus pandemic increased the market prospects for the social robot industry. To deal with the rapid spread of the virus, healthcare sectors employed more robots to support and reduce the risk of the spread of the virus to healthcare workers. Social Assistive Robots (SAR) saw an increase in demand in the retail sector as well. Delivery from social robots increased significantly to ensure quarantine and distance adhering containment measures implemented by the government. Tho contain the spread of the virus, innovation in the robotics industry resulted in the creation of Pre-screening Experience Through Robotic Assessment (PETRA) social robots by Merck Group. PETRA can detect common yet undiagnosed diseases. The social robot has widened the opportunities for the industry and will saw a surge in adoption during the forecasted period.Key Topics Covered: 1. Introduction2. Research Methodology 3. Executive Summary3.1. Research Highlights4. Market Dynamics4.1. Market Drivers4.2. Market Restraints4.3. Porters Five Forces Analysis4.4. Industry Value Chain Analysis5. Social Robots Market, by Application5.1. Introduction5.2. Hardware5.3. Software5.4. Service 6. Social Robots Market, by End Users6.1. Introduction6.2. Healthcare6.3. Education6.4. Retail6.5. Entertainment6.6. IT and communication6.7. Household6.8. Others 7. Social Robots Market, by Geography7.1. Introduction7.2. North America 7.2.1. USA7.2.2. Canada7.2.3. Mexico7.3. South America7.3.1. Brazil7.3.2. Argentina7.3.3. Others 7.4. Europe 7.4.1. Germany7.4.2. France7.4.3. United Kingdom7.4.4. Italy7.4.5. Spain 7.5. Middle East and Africa7.5.1. Saudi Arabia7.5.2. UAE7.5.3. Israel7.5.4. Others 7.6. Asia Pacific7.6.1. China7.6.2. Japan7.6.3. South Korea7.6.4. India7.6.5. Thailand7.6.6. Taiwan7.6.7. Indonesia 7.6.8. Others 8. Competitive Environment and Analysis8.1. Major Players and Strategy Analysis8.2. Emerging Players and Market Lucrative8.3. Mergers, Acquisition, Agreements, and Collaborations8.4. Vendor Competitiveness Matrix9. Company Profiles9.1. Blue Frog Robotics and Buddy9.2. Reach Robotics9.3. Knightscope Inc.9.4. Intuition Robotics9.5. AIST9.6. Furhat Robotics9.7. SoftBank Robotics9.8. Sony9.9. Merck groupFor more information about this report visit https://www.researchandmarkets.com/r/zaflrn Media Contact: Research and Markets Laura Wood, Senior Manager [email protected] For E.S.T Office Hours Call +1-917-300-0470 For U.S./CAN Toll Free Call +1-800-526-8630 For GMT Office Hours Call +353-1-416-8900 U.S. Fax: 646-607-1904 Fax (outside U.S.): +353-1-481-1716 SOURCE Research and Markets http://www.researchandmarkets.com Do not sell or share my personal information: