Can One AI Model Control Every Robot? | Sergey Levine

What will it take to build robots that can truly operate in the real world? 🤖In this episode of SparX, Mukesh Bansal is joined by Sergey Levine, computer scientist and co-founder of Physical Intelligence, to explore the future of AI and robotics and why giving machines physical intelligence is one of the hardest problems in AI.From Moravec’s Paradox and imitation learning to general-purpose robots and robot learning, we explore why AI that can write, reason and talk still struggles with the physical world, and what it will take to build robots that can learn, adapt, and perform a wide range of tasks.Whether you’re an AI enthusiast, robotics researcher, builder, or simply curious about where AI is headed, this conversation offers a glimpse into the next frontier of artificial intelligence.In this episode, we discuss:* Why physical intelligence is so difficult* Moravec’s Paradox and what it tells us about AI* How robots learn from human demonstrations* The race toward general-purpose robots* Why robots struggle to generalize to new environments* What the future of AI-powered robotics could look likeThe next breakthrough in AI will be physical.