Hacker News new | ask | show | jobs
by Ekaros 296 days ago
I think it was and probably still is valid statement on where reinforced learning tend to end to. Something in essence exploiting the physics system.

I think there would be lot less backlash if the end was graceful, smooth and natural looking. But it was not.

1 comments

Getting 3D models to move in a natural way through RL is largely a solved problem. And listen to what they were saying in the video. Whatever they were demoing there, they were also thinking of using for a zombie model in a video game. Ragdolls in video games wasn't a new concept at the time either.

Out of context & blown out of proportion.