Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: ChessCoach – A neural chess engine that comments on each player's moves

chrisbutner.github.io

Post reply on HN
Print view View original

21–30 of 92 posts

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#21

Post by a_t48 »

cbutner wrote: post #10
prezjordan wrote: post #8

Remarkably cool. My only bit of feedback is that it's unfortunate so much of the commentary uses masculine pronouns (I found it distracting). Is that just a consequence of the training data?

Yeah, that's a massive problem with the natural language domain all across machine learning. Unfortunately it's very difficult to track down training data for chess commentary in the first place, let alone trim down biases. For reference, I was able to gather about 1 million samples, but it really needs a billion. Hopefully through data augmentation and better general intelligence models we can make better progress o…

You might be able to kludge a fix to tokenize the output and replace he/him/she/her with them/their. It's not as sexy as the engine outputting the correct words, but it should get the job done.

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#22

Post by deviation »

sobellian wrote: post #20

One of my pet peeves is when folks compare human elo ratings from FIDE to (presumably) engine elo ratings from CCRL, a league for chess engines. They're not the same! This is like comparing points between your local youth soccer league and La Liga. All we need to say is that this engine is only 100 elo behind Stockfish. That is akin to occasionally beating God in a game of chess.

> That is akin to occasionally beating God in a game of chess.

So you're telling me ... There's a chance?

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#24

Post by zeptonix »

zeptonix wrote: post #23

Amazing, incredible idea. Run with it for a bit. Setup a hosted, super easy to use version, that's billed ultimately at cost plus but in an easy to understand way (maybe # of games, that sort of thing).

Responding to myself, but I also think a "why did I lose?" / "where did I first go wrong?" variation would be super popular with people.

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#25

Post by buzzin_ »

sobellian wrote: post #20

One of my pet peeves is when folks compare human elo ratings from FIDE to (presumably) engine elo ratings from CCRL, a league for chess engines. They're not the same! This is like comparing points between your local youth soccer league and La Liga. All we need to say is that this engine is only 100 elo behind Stockfish. That is akin to occasionally beating God in a game of chess.

I thought Alpha Zero is stronger that Stockfish, or was, if it doesn't exist anymore.

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#26

Post by thomasahle »

So do I understand correctly: This is a new head on top of the AlphaZero model?

That is, in addition to the usual evaluation and policy heads, this takes the intermediate board representation and outputs a seed vector that is fed into a transformer text generator?

Or do other things go into the seed? Like the search tree somehow? Otherwise I suppose the commentary will not be able to comment on deeper tactics?

Or maybe this doesn't work using a seed vector at all, but with a custom integration from the board into the transformer somehow?

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#27

Post by modeless »

cbutner wrote: post #2

This took about a year and a half – a little over a year coding in between experiments and training. It's a chess engine with a primary neural network just like AlphaZero or Leela Chess Zero's, but it adds on a secondary "commentary decoder" network based on Transformer architecture to comment on positions and moves. All of the code and data for training and search is from scratch, although it does use Stockfish code…

Where did you source the commentary dataset?

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#28

Post by butwhywhyoh »

prezjordan wrote: post #8

Remarkably cool. My only bit of feedback is that it's unfortunate so much of the commentary uses masculine pronouns (I found it distracting). Is that just a consequence of the training data?

What makes it unfortunate that it used masculine pronouns? If it had used purely female pronouns would that make you feel better?

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#29

Post by taftster »

zeptonix wrote: post #24
zeptonix wrote: post #23

Amazing, incredible idea. Run with it for a bit. Setup a hosted, super easy to use version, that's billed ultimately at cost plus but in an easy to understand way (maybe # of games, that sort of thing).

Responding to myself, but I also think a "why did I lose?" / "where did I first go wrong?" variation would be super popular with people.

Agreed. These two ideas would be killer features! I want to analyze a game, not from where I am at now, but where I was 10 moves ago when everything started to go sideways. I want an AI that can tell me where my instincts or intuitions are off and the reasoning why. e.g. attacking the middle is generally good, but did I focus too much on that and missed some important lines in other places.

Re: Show HN: ChessCoach – A neural chess engine that comments on each player's moves

#30

Post by potatoman22 »

butwhywhyoh wrote: post #28
prezjordan wrote: post #8

Remarkably cool. My only bit of feedback is that it's unfortunate so much of the commentary uses masculine pronouns (I found it distracting). Is that just a consequence of the training data?

What makes it unfortunate that it used masculine pronouns? If it had used purely female pronouns would that make you feel better?

Probably not. The issue with always using one set of pronouns is that people with other pronouns may feel isolated. For example, always assuming software devs are male could cause female devs to feel like they don't belong.

Realistically, it's really hard to fix this encoded bias in language models.

Post reply on HN