In a quite unexpected turn of events, it is claimed that OpenAI’s ChatGPT “got absolutely wrecked on the beginner level” while playing Atari Chess. Citrix Architecture and Delivery specialist, Robert Jr. Caruso, discovered this gameplay skill anomaly over the weekend. Caruso pitted the 1979 Atari Chess title, played within an emulator for the 1977 Atari 2600 console gaming system, against the might of ChatGPT 4o.
A little computer vs Chess history
The concept of computing performance being graded by chess-playing ability is one firmly embedded in nerd lore. Chess computer games were popular from the early days of consoles and home computing, with computing and chess enthusiasts going to great lengths to grade available chess-engine abilities versus a Grandmaster of ‘the game of kings.’
IBM’s Deep Blue supercomputer made history in 1997 when it defeated Garry Kasparov, the reigning world chess champion at the time. Instrumental to its victory, Deep Blue leveraged brute force techniques and evaluated 200 million possible chess moves per second. However, Kasparov struck back after losing the first of the scheduled six chess matches, with the eventual score of 4-2 in his favor.
In 2025, the Deep Blue supercomputer’s processing power of approximately 11.4 GFLOPS seems puny compared to even entry-level modern processors. So, one might expect an Atari Chess running in an almost 48-year-old games console emulation instance to easily be beaten by ChatGPT…
ChatGPT humbled by an Atari 2600
As Caruso points out in his LinkedIn post, the Atari 2600 had very little in the way of computing power. It was powered by a MOS Technology 6507 processor running at 1.19 MHz, and its performance would probably be more sensibly measured in KFLOPS, not GFLOPS. The chess engine that Atari Chess ran only thinks one to two moves ahead, asserts the Citrix Engineer.
Caruso says he tried to make it easy for ChatGPT, he changed the Atari chess piece icons when the chatbot blamed their abstract nature on initial losses. However, making things as clear as he could, ChatGPT “made enough blunders to get laughed out of a 3rd grade chess club,” says the engineer.
Tragically, though ChatGPT 4o kept promising to improve its mastery of the game, the old 8-bit gaming platform would continue to beat it for as long as Caruso had the patience. Despite his direct assistance during the game sessions, ChatGPT couldn’t muster the smarts to beat the Atari Chess 'beginner' opponent and eventually “conceded,” according to this LinkedIn tale.
The news flow regarding artificial intelligence seems to swing between extremes. Sometimes AI can astound with its capabilities, and other times it might be laughable, or even dangerously inadequate. This story can be squarely filed under the latter.
Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.
