An engineer experimenting with ChatGPT found the AI model boasting about its potential to win at chess. The engineer, Robert Caruso, set up a challenge to see how it would fare against an old-school Atari 2600 chess program, which led to ChatGPT’s unexpected defeat.

Caruso explained that chess engines today far surpass human players, but were still astounded by ChatGPT’s performance. “It wanted to find out how quickly it could beat a game that only thinks 1-2 moves ahead on a 1.19 MHz CPU,” he noted on LinkedIn.

Despite being presented with a clear chessboard, ChatGPT struggled, mistaking rooks for bishops and failing to track the pieces’ positions. “It made enough blunders to be laughed out of a 3rd-grade chess club,” Caruso remarked.

ChatGPT suggested it would improve if given another chance, yet after continuous errors, it begrudgingly admitted defeat, stating, “It wished it hadn’t played Atari today.” This laughable defeat shines a light on AI’s limitations, provoking thoughts about its understanding of context and strategies beyond just data processing.