←back to thread

Grandmaster-level chess without search

(github.com)
365 points lawrenceyan | 5 comments | 17 Oct 24 19:13 UTC | HN request time: 0.62s | source
Show context
chvid ◴[18 Oct 24 08:14 UTC] No.41877343[source]▶
But the gigantic synthetic dataset that is used for training is created with plenty of traditional search. So it is all a bit silly but I guess cool none the less ...
replies(3): >>41877359 #>>41877644 #>>41877711 #
1. amunozo ◴[18 Oct 24 09:17 UTC] No.41877644[source]▶
It's a knowledge distillation. You can then use this smaller, more efficient models instead of the larger one.
replies(2): >>41878341 #>>41883075 #
2. chvid ◴[18 Oct 24 11:23 UTC] No.41878341[source]▶
Or maybe it is just memorizing a very large number of games.
replies(2): >>41880894 #>>41882390 #
3. azakai ◴[18 Oct 24 16:24 UTC] No.41880894[source]▶
They address the possibility of memorization in the PDF:

> This effect cannot be explained by memorization since < 1.41% of the initial puzzle board states appear in our training set.

4. tech_ken ◴[18 Oct 24 18:58 UTC] No.41882390[source]▶
Seems more like a 'compression' of the large number of games, or even like an approximate 'index' of the database
5. alkonaut ◴[18 Oct 24 20:18 UTC] No.41883075[source]▶
Is this network smaller than stockfish and by what metric is that?