You may be surprised to learn about the surprising chess matches involving the Atari 2600, a game console that debuted in 1977. This 46-year-old device, featuring a modest 8-bit processor running at just 1.19 MHz and a mere 128 bytes of RAM, has proven to be a formidable opponent against modern AI chatbots like ChatGPT and Microsoft’s Copilot. Recent insights shared by infrastructure architect Robert Caruso revealed that when pitted against ChatGPT, the Atari 2600 triumphed, even at the beginner level. Caruso noted that ChatGPT “got absolutely wrecked,” explaining that it struggled to identify chess pieces correctly, mistaking rooks for bishops and losing track of pieces on the board.
Despite its advanced technology, ChatGPT’s performance was less than impressive, with the chatbot admitting its defeat was due to the obscure icons used by Atari. However, even after switching to standard chess notation, the results were the same. Caruso had to intervene multiple times during a match, resetting ChatGPT’s awareness of the board and ultimately forcing it to concede. In a subsequent matchup, Copilot also underestimated the Atari 2600, believing it could think several moves ahead.
Yet, it faced challenges similar to those encountered by ChatGPT, needing constant board updates. In the end, the Atari 2600 achieved two consecutive victories over Copilot. Curiously, when it came to setting up a match with Gemini, another advanced AI from Google, Gemini opted to cancel rather than risk an embarrassing loss. It acknowledged its limitations in chess-playing capabilities and determined that withdrawing was the most sensible option.
Despite its outdated technology, the Atari 2600’s unexpected success against these sophisticated AI systems raises intriguing questions about the distinctions between different types of artificial intelligence.