Sergey Levine: Current State of Humanoid Robotics, China & Future Predictions

<p>Sergey Levine is one of the world&#39;s top robotics researchers and co-founder of Physical Intelligence. We talked about where humanoid robotics is today, thoughts on the Chinese robotics ecosystem, and his predictions for future timelines.</p><p><br></p><p>• My ergonomic keyboard project I mentioned, you can follow along here: https://read.compose.llc/</p><p>• The Kickstarter page for it: https://www.kickstarter.com/projects/ryanlpeterman/compose-simple-ergonomics-beautifully-done</p><p><br></p><p>Podcast links:</p><p><br></p><p>• YouTube: https://youtu.be/9OSbaPjv0Rc</p><p>• Apple: https://podcasts.apple.com/us/podcast/the-peterman-pod/id1777363835</p><p>• Transcript: https://www.developing.dev/p/sergey-levine-current-state-of-humanoid?r=n49ky</p><p><br></p><p>Thank you to this episode&#39;s sponsor for supporting my work:</p><p><br></p><p>• WorkOS: makes your app Enterprise Ready with easy to use APIs to add SSO, SCIM, RBAC, and more in just a few lines of code, check them out at https://workos.com/</p><p><br></p><p>Timestamps:</p><p><br></p><p>(00:00) Intro</p><p>(00:37) Where are we today</p><p>(04:20) Most surprising capabilities so far</p><p>(07:03) The most inspiring real world robotics</p><p>(08:36) If OpenAI or Anthropic got into robotics</p><p>(10:22) Chinese robotics</p><p>(13:15) Will one lab breakout from the rest</p><p>(16:59) Thoughts on a concrete roadmap</p><p>(21:03) Generalization and demonstrating it</p><p>(26:04) Types of data and which is best for robotics</p><p>(34:34) Why humanoid robotics differs from Waymo</p><p>(37:10) If humanoid robotics failed here is why</p><p>(39:55) Are there hot take modeling architectures in robotics</p><p>(42:05) Thoughts on AI safety in robotics</p><p>(46:44) Top robotics research paper recommendation</p><p>(49:35) Why is Boston Dynamics less top of mind</p><p>(53:47) Advice for his younger self</p><p>(56:42) Outro</p><p><br></p><p>Where to find Sergey:</p><p><br></p><p>• Google Scholar: https://scholar.google.com/citations?user=8R35rCwAAAAJ&amp;hl=en</p><p>• Website: https://people.eecs.berkeley.edu/~svlevine/</p><p>• Wikipedia: https://en.wikipedia.org/wiki/Sergey_Levine</p><p>• X/Twitter: https://x.com/svlevine?lang=en</p><p>• LinkedIn: https://www.linkedin.com/in/sergey-levine-5a31a24/</p><p><br></p><p>Where to find Ryan:</p><p><br></p><p>• Newsletter: https://www.developing.dev/</p><p>• X/Twitter: https://x.com/ryanlpeterman</p><p>• LinkedIn: https://www.linkedin.com/in/ryanlpeterman/</p><p>• Threads: https://www.threads.com/@ryanlpeterman</p><p>• Instagram: https://www.instagram.com/ryanlpeterman</p><p>• TikTok: https://www.tiktok.com/@ryanlpeterman</p><p><br></p><p>Referenced in this episode:</p><p><br></p><p>• Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware (ALOHA / ACT paper): https://arxiv.org/abs/2304.13705</p><p>• Emergence of Human to Robot Transfer in Vision-Language-Action Models: https://arxiv.org/abs/2512.22414</p><p>• Summary of human to robot paper: https://www.pi.website/research/human_to_robot</p>