You can't read every day about ChatGpt, who knows everything about Openai, being taken to the cleaner by the 46-year-old 8-bit program Star Wars-era Chess Engine.
But that's exactly what happened in a matchup organized by Citrix software engineer Robert Caruso.
ChatGpt, Atari Match Up
It all began as a casual conversation between Caruso and ChatGpt about using AI in chess. Following on, the cheeky chatbot proposed playing chess against the 1977 vintage Atari 2600.
The Stella emulator's 90-minute error comedy was then made to make repeated stupid and illegal moves, as ChatGpt was causing confusion for the bishop, forgetting where the pieces were.
Switching the interface to standard algebraic notation – Caruso thought this would give Chatgup a clearer view of the game – also useless, and Atari won fair and square.
The co-pilot loses, and Gemini retreats
Atari's next opponent, Microsoft's Copilot, despite the bot's tall claims, did not claim that it could “advance 10-15 moves” and could easily win. It was “Game Over” that Copilot lost two pawns, knights and bishops by his seventh turn, and transferred the Queen to direct capture.
Google's Gemini AI turns out to be the smartest thing in the lot after reviewing Caruso's previous matches. It refused to take on Atari and cited “time efficiency” and “wise decisions” as reasons for pulling out.
LLMS Limitations
Atari's victory over some of the world's most advanced and large-scale language models is not only shocking, but also speaks.
In an age of AI personification, the loss of billions of dollar chatbots to the 4kb chess engine shows that these tools design linguistic predictions only, rather than structured reasoning.
Image credit: AlexBuess/Shutterstock
