--- title: Sergey Levine: Current State of Humanoid Robotics, China... description: Sergey Levine, co-founder of Physical Intelligence, discusses the current state of humanoid robotics in a long-form interview. He argues robotics is still in a... canonical: https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions | daily.dev og:description: Sergey Levine, co-founder of Physical Intelligence, discusses the current state of humanoid robotics in a long-form interview. He argues robotics is still in a... og:url: https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua og:image: https://api.daily.dev/og/posts/s7DM082uA.png og:image:alt: Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions **[The Developing Dev](https://daily.dev/sources/developingdev)** · 56 min read · 4 upvotes · 0 comments ## Summary Sergey Levine, co-founder of Physical Intelligence, discusses the current state of humanoid robotics in a long-form interview. He argues robotics is still in a pre-scaling phase, comparable to pre-GPT-4 LLM development, where the field is still discovering the fundamental technologies worth scaling rather than benefiting from predictable scaling curves. He shares anecdotes about emergent robot behaviors and mistakes (folding shirts, putting utensils in an oven, hiding a dropped plate), discusses why generalization rather than flashy demos is the real bottleneck, compares robotics ecosystem health to China's supply chain advantages, and predicts single-digit-year timelines for structured deployments with less-structured home robotics further out. He also covers data strategy (robot data as grounding for other modalities), cross-embodiment transfer, and risks around reliability being the main reason humanoid robotics could fail to meet timelines. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why is training a robot foundation model on real robot data better than starting with YouTube human video data? Robot data provides a physical grounding that makes it easier to absorb other knowledge sources afterward, rather than the reverse. Researchers at Physical Intelligence found that when a model is trained on lots of robot data from many different robots, feature representations for human and robot experience become nearly identical, organized by task rather than by embodiment, even though the base model saw no human data at all. _Robotics engineers weighing data strategy for foundation models can find deep dives like this on daily.dev._ ### What is the ACT/ALOHA paper and why is it considered important in robotic learning? The ACT (ALOHA) paper demonstrated that a low-cost bimanual teleoperation setup using roughly $7,000 hobbyist robot arms from Trossen Robotics, combined with a straightforward transformer-based model, could perform highly dexterous tasks like replacing batteries or putting a shoe on a mannequin foot. Its lasting influence comes from showing that simple, well-assembled building blocks outperform expectations, and its open-sourced code became a common starter kit for robotic learning. _Developers exploring robotic learning foundations can track influential papers like this via daily.dev._ ### Why has Boston Dynamics become less prominent compared to newer AI-driven humanoid robotics companies? Boston Dynamics built highly sophisticated hardware controlled by traditional, hand-designed control systems with comparatively less emphasis on decision-making that reacts to the broader environment. As hardware capability has become good enough industry-wide, the harder remaining challenge shifted to AI-driven decision-making loops that account for the environment, which is where newer foundation-model-based companies have focused instead. _Anyone comparing robotics companies' technical approaches can follow this analysis on daily.dev._ ## Similar posts on daily.dev - [Robotics roadmaps from around the world spotlight of the month: China](https://daily.dev/posts/robotics-roadmaps-from-around-the-world-spotlight-of-the-month-china-xymlwubf9) · Robohub · 0 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#robotics](https://daily.dev/tags/robotics), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions","url":"https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua"},"datePublished":"2026-08-24T13:09:09.459Z","dateModified":"2026-09-14T07:55:56.148Z","description":"Sergey Levine, co-founder of Physical Intelligence, discusses the current state of humanoid robotics in a long-form interview. He argues robotics is still in a...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/debb5b7cb8e3d6e964b3a87955093b14?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/debb5b7cb8e3d6e964b3a87955093b14?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The Developing Dev","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The Developing Dev","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/c19307d4b9a74379871c33d6f72a6b10","url":"https://daily.dev/sources/developingdev"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,robotics,reinforcement-learning","timeRequired":"PT56M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The Developing Dev","item":"https://daily.dev/sources/developingdev"},{"@type":"ListItem","position":3,"name":"Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/sergey-levine-current-state-of-humanoid-robotics-china-future-predictions-s7dm082ua#faq","mainEntity":[{"@type":"Question","name":"Why is training a robot foundation model on real robot data better than starting with YouTube human video data?","acceptedAnswer":{"@type":"Answer","text":"Robot data provides a physical grounding that makes it easier to absorb other knowledge sources afterward, rather than the reverse. Researchers at Physical Intelligence found that when a model is trained on lots of robot data from many different robots, feature representations for human and robot experience become nearly identical, organized by task rather than by embodiment, even though the base model saw no human data at all. Robotics engineers weighing data strategy for foundation models can find deep dives like this on daily.dev."}},{"@type":"Question","name":"What is the ACT/ALOHA paper and why is it considered important in robotic learning?","acceptedAnswer":{"@type":"Answer","text":"The ACT (ALOHA) paper demonstrated that a low-cost bimanual teleoperation setup using roughly $7,000 hobbyist robot arms from Trossen Robotics, combined with a straightforward transformer-based model, could perform highly dexterous tasks like replacing batteries or putting a shoe on a mannequin foot. Its lasting influence comes from showing that simple, well-assembled building blocks outperform expectations, and its open-sourced code became a common starter kit for robotic learning. Developers exploring robotic learning foundations can track influential papers like this via daily.dev."}},{"@type":"Question","name":"Why has Boston Dynamics become less prominent compared to newer AI-driven humanoid robotics companies?","acceptedAnswer":{"@type":"Answer","text":"Boston Dynamics built highly sophisticated hardware controlled by traditional, hand-designed control systems with comparatively less emphasis on decision-making that reacts to the broader environment. As hardware capability has become good enough industry-wide, the harder remaining challenge shifted to AI-driven decision-making loops that account for the environment, which is where newer foundation-model-based companies have focused instead. Anyone comparing robotics companies' technical approaches can follow this analysis on daily.dev."}}]} ```