Back to Y Combinator

Chelsea Finn: This is the State of the Art in Robotics

Y CombinatorAugust 12, 202658m
Topics42
Physical Intelligence and General-Purpose Robotics0:06Recent Robot Capabilities1:00General-Purpose Models and Real-World Impact1:31Historical Timeline of AI in Production2:00Key Differences Between Digital AI and Physical AI3:30Waymo's Autonomous Achievement5:02Long-Term Autonomy Requirements5:31Iterative Model Improvement Process7:00Self-Improving AI Systems7:30Reinforcement Learning Challenges in Robotics8:00Addressing Dead-End Trajectories9:30Value Function Amortization11:00Multi-Stage Improvement Algorithm13:00Espresso-Making Demonstration13:30Additional Real-World Applications15:00Quantitative Performance Measurements16:01Long-Term Autonomy Achievements16:30Memory Requirements for Complex Tasks17:31Memory Implementation Challenges18:30Multi-Time-Scale Memory System19:31Kitchen Cleaning Task Demonstration20:30General-Purpose Model Development21:31Compositional Generalization Milestone23:30Developing General Purpose Robot Models24:50Training Recipe for General Purpose Models27:01Model Deployment and High-Level Policies29:01Performance Comparison with Specialist Models30:30Compositional Generalization Testing31:30Quantitative Results and Ablation Studies35:01Current State and Real-World Deployment37:01Q&A: ChatGPT Moment for Robotics39:31Transitioning to Generalist Policies41:01PhD Considerations for Industry Careers42:30Robotics Data Equivalent to Internet-Scale Training45:31Open Source Democratization of Robotics Models48:30Open Source vs Closed Source Models in Robotics48:48Robot Control Architecture49:30Imagination and Future Prediction in Robotics51:30Improving Robot Speed53:30Emergent Robot Capabilities54:30Future Directions and Reliability56:00Breaking Into Robotics from Software Engineering56:30
In a Nutshell

Chelsea Finn's Physical Intelligence has developed general-purpose robotics models that achieve 90%+ reliability on real-world tasks like espresso-making and kitchen cleaning through a self-improving system combining foundation model pre-training, human intervention for dead-end avoidance, and reinforcement learning using a value function trained on diverse robot experience. The PIO7 model demonstrates out-of-the-box performance matching or exceeding task-specific fine-tuned specialists, with compositional generalization across robot platforms and objects never seen in training data. The approach enables long-term autonomous operation (13+ hours) without human supervision by implementing multi-time-scale memory systems that compress extended task histories into actionable text summaries.

AI-Generated Notes

These notes were generated by AI and may contain inaccuracies.

Chelsea Finn founded Physical Intelligence two years ago to develop robots capable of performing any task in the real world. One year prior, she presented progress on complex tasks including unloading and folding laundry, and demonstrated robots successfully completing useful tasks in previously unseen environments.

Since the previous presentation, robots have achieved additional tasks including washing a greasy pan, peeling a carrot, making a grilled cheese sandwich, and slicing a zucchini. The focus shifts from showcasing individual capabilities to understanding what enables robots to become useful in real-world applications.

Two critical aspects must be addressed: developing general-purpose models and bringing these models into the real world for actual impact. The discussion begins with the challenges of real-world deployment.

A timeline of major production launches leveraging machine learning reveals key milestones. Early applications included product recommendations and ad ranking. Five years later, deep learning emerged for similar applications, offering advantages because the algorithm could be applied out-of-the-box to complex input-output scenarios. The pivotal moment occurred in 2022 with ChatGPT, the first general-purpose model used by many people in the real world, reaching one million users within five days. More recently, coding agents like Cursor have demonstrated real-world utility.

Sign in to read the full notes

Get access to AI-generated notes, topic timestamps, and more.