Episode Details

Robotics may be on the verge of a breakthrough—but only if robots can learn to adapt in the messy real world. In this condensed version of Possible Podcast, Reid Hoffman talks with Chelsea Finn, co-founder of Physical Intelligence and a Stanford researcher in meta-learning, about why general-purpose robotics is so hard, why real-world data matters, and how tokenized actions plus a diffusion head improved instruction following. Learn how their team went from months of failing at laundry folding to robots that can recover from mistakes, open a fridge, and handle changing environments in warehouses and homes. This quick listen distills a full-length conversation into a time-saving summary that captures the key ideas on artificial intelligence, machine learning, business model innovation, productivity, and the ethics of AI alignment. Listen now to get the key ideas in minutes.