Why Bad Data Makes Better Robots @reidhoffman
Why Bad Data Makes Better Robots  @reidhoffman
Uploaded September 2026 | Updated November 2026, 3 hours ago
Useful robots won’t be programmed one task at a time; they’ll need to adapt to unfamiliar objects, environments, and robot bodies. Physical Intelligence cofounder Chelsea Finn (named last month to the TIME100 AI list for 2026) joins Reid Hoffman and Aria Finger to explain how general-purpose robot models learn to act in the messy physical world. She explores why lower-quality training data can strengthen a model, what months of failed laundry-folding attempts revealed, and how the same system can work across different robot platforms. A robot mistaking an oven for a drawer becomes a lesson in the strange line between useful generalization and obvious error. They also examine why robot demos can mislead, how hardware reliability can stall progress, and what it will take to move AI off the screen and into the world.

Congratulations to Chelsea, recently named one of TIME’s 100 most influential people in AI:
time.com/collection/time100-ai/2026/chelsea-finn

For more info on the podcast and transcripts of all the episodes, visit https://www.possible.fm/podcast/


#ai #technology #business #podcast
Why Bad Data Makes Better RobotsWhat the 2026 midterm results really mean | Reid HoffmanThe Experiment that made $4 BILLION (W/ Creators of Cryptopunks)Who should govern AI?Microsofts CEOs MUST READ BookCould AI end one-size-fits-all learning?You need to embrace the technical velocity of AI.Were living in a world after virtue right now but darker times than this have been survived before.Do CEOs really understand AI?AI could escalate cyber warfareThe AI Energy War Is HereAmericas Aerospace Rebirth
Reid Hoffman |

Why Bad Data Makes Better Robots

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER