A breakthrough post-training result for physical AI via self-play: a strong base model like S1 can learn to complete extremely dexterous and dynamic tasks, like soccer, by competing against itself in a simulation.

4 points•mostlyk•7 days ago•1 comment•

1 comment

rvz7 days ago
Looks impressive. A breath of fresh air.

Read the full thread on Hacker News →

Related stories