PlatPhorm Podcasts

Chelsea Finn: Building Robots That Can Do Anything

Y Combinator Startup Podcast · July 22, 2025

Chelsea Finn: Building Robots That Can Do Anything artwork

episode · Checking source

Chelsea Finn: Building Robots That Can Do Anything

Y Combinator Startup Podcast

Chelsea Finn on June 17th, 2025 at AI Startup School in San Francisco.From MIT through her PhD at Berkeley, where she pioneered meta‑learning methods, and Google Brain, Chelsea Finn has built her career around teaching machines how to learn. Now an Assistant Professor at Stanford and co‑founder of Physical Intelligence, she’s using that foundation to bring learning-driven robotics into messy, real-world environments rather than confined lab setups.In this talk, Chelsea traces the evolution of her team’s work—from early experiments on robotic grasping and vision to today’s ambitious efforts at folding laundry, tidying kitchens, and generalizing across tasks—all without hand-crafted code. Instead, they used scalable foundation models and massive datasets, teaching robots physical common sense as they learn by doing. She shares stories of the rocky setbacks, the surprises hidden in data, and the moment it all clicked: robots equipped with generalizable physical intelligence can indeed adapt and assist in the unpredictable world around us.

View original

Chelsea Finn on June 17th, 2025 at AI Startup School in San Francisco.From MIT through her PhD at Berkeley, where she pioneered meta‑learning methods, and Google Brain, Chelsea Finn has built her career around teaching machines how to learn. Now an Assistant Professor at Stanford and co‑founder of Physical Intelligence, she’s using that foundation to bring learning-driven robotics into messy, real-world environments rather than confined lab setups.In this talk, Chelsea traces the evolution of her team’s work—from early experiments on robotic grasping and vision to today’s ambitious efforts at folding laundry, tidying kitchens, and generalizing across tasks—all without hand-crafted code. Instead, they used scalable foundation models and massive datasets, teaching robots physical common sense as they learn by doing. She shares stories of the rocky setbacks, the surprises hidden in data, and the moment it all clicked: robots equipped with generalizable physical intelligence can indeed adapt and assist in the unpredictable world around us.

Published
July 22, 2025
Status
active
GUID hash
c551cd18fa5ad628e39fc5a672bd3e786dba95a4556701744d81e826beef63c5
Archive key
chelsea-finn-building-robots-that-can-do-anything--entry_47976651d067213b1862a0277892
Archive id
entry_47976651d067213b1862a0277892