2024-12-22 at 17:38 more notes on Parables on the Power of Planning in AI (noam brown)

And this was, at the time, state of the art for predicting human moves in chess.
29:48
Now, one thing that's really interesting about MAIA is that for high Elo models, it was about 100 to 300 points--
29:56
Elo points-- below the target Elo rating. So if you were to train it on 2,000 Elo-rated humans,
30:03
it would only be about 1,700 Elo. For the lower Elo ratings, this ended up not being a problem.
30:08
For the higher Elo ratings, it was a challenge. Now, one hypothesis for why this is the case
30:16
is that approximating human planning is hard for neural nets.
30:21
When you're at 2,000 Elo, your planning process is so sophisticated that it's actually quite difficult
30:28
to distill that down into a neural net. And I think one piece of evidence for this is that there is one version of chess where there was
30:37
no Elo gap for the MAIA model, and that was bullet chess, where humans have very, very little time to plan ahead
30:43
and they just have to basically act on instinct.

emphasis on:
When you're at 2,000 Elo, your planning process is so sophisticated that it's actually quite difficult to distill that down into a neural net. And I think one piece of evidence for this is that there is one version of chess where there was no Elo gap for the MAIA model, and that was bullet chess, where humans have very, very little time to plan ahead and they just have to basically act on instinct.

===

he talks about an openai paper about . verify each step
this is exactly what we do as humans