I Watched a Robot Do 3 Tasks It Was Never Trained On.

A robot just folded laundry, made coffee, and assembled a box. None of those tasks were in its training data.

That's Physical Intelligence's π0.7 — a foundation model for robots that does something new: compositional generalization. It learns individual skills like pick up, fold, and place separately, then combines them on its own to solve tasks it's never seen.

This is the same leap language models made when they went from memorizing answers to reasoning about new ones. Except now it's happening in the physical world.

Manufacturing, warehousing, and logistics — the implications are enormous.

Build. Ship. Scale. Repeat.
🔔 Subscribe @NeuralNeeraj for daily AI deep-dives.

#Robotics #AI #PhysicalIntelligence #FutureOfWork #AGI #Shorts

Видео I Watched a Robot Do 3 Tasks It Was Never Trained On. канала Neural Neeraj