If Russian novelist and Crimean War veteran Leo Tolstoy is right about war being a chess match, we may not have to worry so much about fighting AI-led robots on a futuristic, “Terminator” like battlefield.
At least, not yet.
It turns out that some AI models are really bad at chess, specifically ChatGPT. Technology site Futurism found software engineer Robert Caruso’s LinkedIn post, wherein he reported that the popular AI bot got “absolutely wrecked” by an Atari 2600 playing Atari Chess.
“Video Chess: Special Edition”
There was nothing special about the game itself, “Atari Chess,” which was released in 1979 for the Atari 2600 game console. A 4k game at a time when most games were just 2k, the game was one of the most groundbreaking advances in video games at the time. It had eight different difficulty levels, which meant the computer might take a considerable amount of time to make its next move.
By today’s standards, the game is extremely primitive, especially compared to open-source chess engines like Stockfish, which can process millions of moves in seconds. “Video Chess” was also known for having no opening move strategy. It also makes some bizarre moves, such as random “king walks” (traveling along its side of the board), various tactical mistakes, and sacrificing pieces for no gain.
And yet, ChatGPT wasn’t even playing on an advanced level.
Atari’s Checkmate
“ChatGPT got absolutely wrecked on the beginner level,” Caruso wrote. “This was after a conversation we had regarding the history of AI in Chess which led to it volunteering to play Atari Chess. It wanted to find out how quickly it could beat a game that only thinks 1-2 moves ahead.”
In multiple attempts over 90 minutes, the AI model struggled to remember which pieces were which and where they would be placed on the chessboard. It also misread the moves of its elderly opponent, blaming its performance (or lack thereof) on the Atari game’s digital art. Like a child, the AI kept asking for a do-over, a chance to perform better if it could just restart the game. But that changed nothing for its 8-bit counterpart.
“Atari’s humble 8-bit engine just did its thing,” he said. “No language model. No flash. Just brute-force board evaluation and 1977 stubbornness.”
Sadly, the engineer did not record the AI’s struggle to play a chess match, but assured us ChatGPT “made enough blunders to get laughed out of a 3rd grade chess club.”
Although ChatGPT is capable of understanding the game of chess, it is still unable to devise its own chess strategy. The Atari 2600 playing “Video Chess” is, on the other hand, the right tool for the job. Atari Chess only plays chess. Like most things in life, knowledge simply cannot make up for ability.
When asked about such a loss, Chinese ChatGPT rival DeepSeek said, “Defeating the Atari 2600 at chess would be trivial for me. It represents a fascinating early step in computer chess, but its playing strength is extremely low compared to anything developed in the last 40+ years. It would be like asking if a Formula 1 car could beat a Model T Ford in a race!”
Don’t Miss the Best of Mighty News
We Are The Mighty is a celebration of military service, with a mission to entertain, inform, and inspire those who serve and those who support them. We are made by and for current service members, veterans, spouses, family members, and civilians who want to be part of this community. Keep up with the best in military culture and entertainment: subscribe to the We Are The Mighty newsletter.