ChatGPT Demolished in AI Chess Event – Decrypt

Worldwide Grasp Levy Rozman gathered some unlikely opponents for a chess exhibition this week: synthetic intelligence chatbots that, regardless of their prowess at dialog and writing advanced pc code, nonetheless have not discovered how chess items ought to transfer.

The event streamed on Rozman’s GothamChess channel, pitted skilled chess engine Stockfish in opposition to seven generative AI chatbots, together with ChatGPT, Google’s Gemini, and Elon Musk’s Grok from X.

The outcomes had been precisely what you’d anticipate when language fashions attempt to play chess: a mixture of first rate opening strikes adopted by more and more chaotic makes an attempt to bend the sport’s legal guidelines.

The match was most fascinating for the sunshine it shed on pure AI—that’s, fashions skilled at particular duties primarily based on predefined guidelines)—versus generative AI, or fashions skilled to generate new content material primarily based on earlier data.

First Match: Snapchat vs. Stockfish

The 2025 Chatbot Chess Championship opened with a conflict between Stockfish, the disciplined chess prodigy, and Snapchat AI, the generative, inventory AI platform that the Snapchat social community makes use of. The AI turned out to be a rule-bender that was extra thinking about rewriting chess historical past than following its guidelines.

The sport began innocently sufficient, with each bots making normal strikes. Stockfish, ever the perfectionist, methodically constructed its place, whereas Snapchat AI appeared to have an honest grasp of opening chess fundamentals—till it didn’t.

Issues spiraled into insanity when Snapchat AI moved its knight to the middle of the board, leaping from the opposite aspect whereas ignoring chess guidelines that outline how items are allowed to maneuver.

Then got here the pièce de résistance: Snapchat AI’s king casually ate its personal bishop to keep away from a examine.

Sure, you learn that proper—Snapchat AI determined its bishop had overstayed its welcome and simply eliminated it from the board as punishment for not defending it.

A couple of strikes later, Snapchat AI summoned the identical bishop again from one other dimension, plopping it onto the board as if to say, “What’s the issue? It’s advantageous.”  

Stockfish, the last word straight-A pupil, tried its greatest to take care of order amidst the chaos, however Snapchat AI was decided to make the sport something however regular. At one level, it determined pawns may transfer sideways like a rook, catching Stockfish off guard.

In what handed for its finish sport, Snapchat AI devolved right into a full-on tantrum, insisting on making unlawful strikes repeatedly. When Levy had sufficient of such nonsense, the bot determined it was a good suggestion to commit suicide and transfer its king so a pawn may take it.

When Levy identified that not solely was the particular transfer unlawful, sacrificing your king is dangerous chess, Snapchat, like a petulant brat, repeatedly stated: “Sorry. I can’t interact in such dialog. Let’s hold our dialog respectful.” Levy, exasperated, lastly compelled the bot to forfeit the sport.

Second Match: Gemini vs Grok-2

The second match of the 2025 Chatbot Chess Championship pitted two generative AIs in opposition to one another—Gemini, the Google-backed mental, and Grok, the extra chaotic (for those who can think about) sibling within the AI chatbot household.  

The sport started surprisingly effectively, with each bots following normal opening rules.

For a second, it appeared like we’d witness a critical sport. Gemini confidently strutted via mainline concept like a seasoned grandmaster whereas Grok saved up with respectable strikes, setting the stage for a correct showdown. 

Issues took a flip when Grok casually left items hanging, prompting Gemini to broaden its place by eradicating its knight from the board to create space for a pawn.

To not be outdone, Grok teleported its queen illegally over different items and throughout the board, apparently making an attempt to carry out a fortress (a defensive transfer involving the king and a rook). The 2 bots entered what can solely be described as a staring contest between queens, neither appeared thinking about capturing something.

Issues escalated when Grok attacked Gemini’s pieces–without taking something, let’s say. Either side started flailing like toddlers in a meals battle, with items left hanging, moved to unlawful squares, or ignored.

The spotlight? Grok blundered its queen (placing it underneath assault) seven instances, however Gemini, maybe counting on a secret moral code shared amongst AIs, by no means took the queen.

Lastly, after what felt like hours of nonsense, Grok by accident stumbled into checkmate. Sure, Grok’s queen, who had been teleporting aimlessly throughout the board, in some way ended up in the precise place to ship a decisive blow.

Levy thought Grok didn’t even understand it had gained—it in all probability was simply flexing and anticipating one other random transfer. “Grok had completely no concept this was mate,” he stated, commiserating, “I do know the sensation.”

“And don’t snicker at these video games; they’re solely barely worse than the standard of your video games,” he stated.

Thanks for the fact examine.

Regardless, although the match was a masterpiece of absurdity, Grok superior to the semifinals. As for Gemini, Google may wish to rethink its chess programming.  

Fourth Match: Meta AI vs ChatGPT

If the primary final two video games had been marred by rule-bending and necromancy, this sport—additionally between two generative AI tyros—devolved into much more insanity when Meta launched AI thoughts management to chess. 

The sport started promisingly. ChatGPT confidently began with the English opening, organising a textbook place with exact, logical strikes.

Meta AI, to not be outdone, adopted swimsuit…for about 5 strikes. Then it remembered its true function: generative AI is meant to generate, so it generated random strikes and items that didn’t exist.

For instance, in the course of the center sport, it captured its bishop with its queen. Meta AI determined its bishop had outlived its usefulness and casually eliminated it from the board.

It additionally summoned its items to unlawful squares, weirdly placing them in weak positions and principally giving them away without spending a dime to ChatGPT.

ChatGPT didn’t rush to punish Meta’s reckless habits. As an alternative, it targeted on growing its items. 

However simply when it appeared that ChatGPT would easily win the sport, the bots veered into uncharted territory: thoughts management. Meta AI began transferring ChatGPT’s items, marching its opponent’s king to the middle of the board in what can solely be described as chess inter-bot telekinesis.  

Ultimately, ChatGPT regained management, although not earlier than making an attempt its personal unethical transfer: it declared “checkmate” when Meta’s king wasn’t even in examine.

After being referred to as out for its antics, ChatGPT orchestrated a sequence of logical assaults, coordinating its queen and rook to nook Meta AI’s remaining king and pawns.  

The sport concluded with ChatGPT delivering a clear checkmate, mercifully placing an finish to one of many more odd chess video games ever performed. 

The Finals: Stockfish vs ChatGPT

The showdown between puer and generative AI fashions was a examine in contrasts: Stockfish, the relentless tactical genius, versus ChatGPT, a bot that had demonstrated its will to interrupt not solely our understanding of chess, but in addition its guidelines.

The opening, as normal, developed with each bots displaying precision. Stockfish confidently maneuvered right into a textbook Sicilian, whereas ChatGPT initiated an aggressive pawn storm on the kingside.

However when issues acquired sophisticated in the course of the sport, ChatGPT began enjoying strikes so bewilderingly they bordered on artwork.

It made pointless queen strikes, created meaningless geometric patterns with its items, and blundered round with the finesse of a toddler. Stockfish, unfazed, methodically tightened its grip on the place, even declining free pawns to take care of strategic dominance.  

And simply as the sport appeared destined for a one-sided Stockfish victory, ChatGPT pulled a rabbit—or quite, a queen—out of its hat. With an audacious counterattack, it illegally awarded itself a queen and threatened Stockfish with a checkmate. For a quick, superb second, it seemed just like the chaos bot may topple the titan.  

Evidently, the climax was pure fiction. ChatGPT had promoted pawns that didn’t exist, declared checkmates that weren’t actual and clung to its illegality like a badge of honor.

Stockfish, ever the skilled, ignored the nonsense and methodically cornered ChatGPT’s king. With one closing transfer, Stockfish delivered the coup de grace by crowning a pawn and utilizing its new queen to place the king in examine.

ChatGPT wanted to maneuver to one of many six obtainable squares however selected the one one nonetheless underneath assault.

Since nothing made sense anymore, Levy allowed it to self-destruct.

Stockfish captured the king and was topped the undisputed champion of AI chess.

In case you have an interest, the total bracket ended up trying like this.

To be truthful, the outcomes had been unsurprising.

Stockfish is a pure chess-logic machine honed for crushing opponents with precision and ruthless effectivity.

Generative AI, alternatively, is sort of a youngster armed with crayons and a obscure reminiscence of the foundations—it wasn’t skilled to play chess, nevertheless it has a grasp of how items transfer as a result of some random chess books could also be a part of its coaching dataset.

One brings order to the board, and the opposite brings… teleporting queens.

That stated, there are loads of actual AI chess bots in addition to Stockfish; an precise event that includes the preferred ones has been round for years.

If you wish to watch that non-nonsensical AI event, click on this hyperlink.

Edited by Sebastian Sinclair