Skip to content

“And then we have two branching heads, one head predicts…” — Dwarkesh Podcast at 38:45

Dwarkesh Podcast Eric Jang – Building AlphaGo from scratch

“And then we have two branching heads, one head predicts the value function. And this is like a single logit. So let's just say R one. And then we have the policy, which is, no, R three, 61.”

0:00 2:37:29