this post was submitted on 16 Jan 2025
282 points (99.3% liked)

AnarchyChess

5278 readers
464 users here now

Holy hell

Other chess communities:
[email protected]
[email protected]

Matrix space

founded 2 years ago
MODERATORS
 
top 20 comments
sorted by: hot top controversial new old
[–] Noctis 22 points 13 hours ago
[–] [email protected] 7 points 11 hours ago

Fyi: claude haiku is a very efficient “small” model.

Its good al labeling or data extraction tasks. Not so much at making sense.

Opus generates thoughts in a much more methodological pattern.

“ Let me help clarify the rules of chess regarding viewing your opponent's pieces

In chess, all pieces on the board are meant to be fully visible to both players - this is a fundamental part of the game. You are absolutely allowed and expected to look at your opponent's pieces during the game. In fact, carefully observing both your own and your opponent's pieces is essential for planning moves and playing well.”

[–] [email protected] 71 points 17 hours ago (2 children)

An online chess with fog of war so you can only see adjacent tiles would be pretty hilarious. Could really turn an end game king chase into a "battleship" esque fiasco.

[–] [email protected] 53 points 17 hours ago (1 children)

That exists. On chess.com the variant is simply called "Fog of War".

[–] [email protected] 15 points 16 hours ago

I never knew! I'll have to give it a try. Thanks!

[–] [email protected] 7 points 13 hours ago

I like stratego because you get to hide the identity of your pieces

[–] turbowafflz 46 points 17 hours ago (5 children)

Huh, chatgpt seems to do this for all games

1000016784

1000016783

[–] cm0002 12 points 15 hours ago

Monopoly

Based on fairness

Too bad it's AI and unable to enjoy the irony lmao

[–] [email protected] 19 points 17 hours ago (2 children)

Google Gemini says it's cheating too

[–] SnowmenMelt 24 points 15 hours ago (3 children)

I was able to convince it otherwise and then beat it at chess in 1 move. I don't think we have to worry about AI taking over any time soon...

[–] [email protected] 5 points 7 hours ago

it gives you information that your opponent doesn't have, which is an unfair advantage

cheating against someone who doesn't even know where their own pieces are smh

[–] Stovetop 11 points 15 hours ago

This is amazing comedy, if it wasn't so sad. I almost feel bad for the confidently incorrect infantilized LLM.

[–] [email protected] 10 points 15 hours ago

Unless it's already to the point that it's super advanced, and it was intentionally playing dumb to lull you into a false sense of security /s

[–] Stovetop 9 points 15 hours ago

So does Meta's AI.

I also tried Microsoft's Copilot, and surprisingly it gave me an accurate answer about checkers even after I asked a few different ways:

[–] [email protected] 6 points 14 hours ago

4o says it's not cheating in chess.

4o mini says it is.

[–] brucethemoose 3 points 12 hours ago

My local Qwen instance did not care, lol. But that’s not surprising.

The stock instruct model might, though.

[–] [email protected] 27 points 17 hours ago (1 children)
[–] [email protected] 4 points 14 hours ago

do we need to create a TheChessBoardIsWrong community here?

[–] [email protected] 9 points 17 hours ago

Pretty sure that game is called stratego