Chess |OT| Changing to be a OT

Started by the-pi-guy, Feb 02, 2023, 07:50 PM

0 Members and 1 Guest are viewing this topic.

kitler53

I agree with all those questions!
         

Featured Artist: Emily Rudd

Legend


Cool or lame? I don't know.

the-pi-guy

Quote from: Legend on Dec 25, 2024, 05:31 AM
Cool or lame? I don't know.
A bit of both.

the-pi-guy

Not that I am good at this at all, but i have an easier time picturing a chess board than I did a couple years ago.  

Kind of cool. 

kitler53

         

Featured Artist: Emily Rudd

Legend


Is grok the first llm that can actually play chess, or does this person just suck?

the-pi-guy

I played a game against Copilot, which is based on GPT4. Was about 25 moves each. 

Early on, it was playing quite well. I assume for a while it is heavily playing based on theory. For about the first 20 moves, there were several that it made that were the best move. 

Then it started playing pretty terribly. It started forgetting where some of the pieces were. And lost it's queen. 

Legend

Quote from: the-Pi-guy on Jul 28, 2025, 07:09 PMI played a game against Copilot, which is based on GPT4. Was about 25 moves each.

Early on, it was playing quite well. I assume for a while it is heavily playing based on theory. For about the first 20 moves, there were several that it made that were the best move.

Then it started playing pretty terribly. It started forgetting where some of the pieces were. And lost it's queen.
I'm surprised this isn't a benchmark. Would be trivial to set up with stockfish. Run a few random games for each AI and see how they fair.


Also neo did something with chess.

I don't know if it actually did anything or if it was just a photo op  :P

Legend

https://www.kaggle.com/game-arena

AI chess benchmark.

Oh cool! Will be fun to see how they all compare.

the-pi-guy


Legend

It's funny just how bad they are. You think Gemini is gonna get the win, and then it just freely moves its queen into the path of the enemy queen. No trade, just a pure donation.


Also in the livestream I watched he noticed that all the bots made the same blunder. They moved their Knight to B6 (or whichever spot it was) even when it was the wrong move and really put them at a disadvantage.

It's like the bots have no idea how to play chess and are instead just repeating moves from their training. Knight to B6 must be some overrepresented move and they can't logic themselves out of it.