How an old Atari console forced modern AI to surrender without a fight
The super-powerful Google Gemini refused to play chess with an Atari console from 1977. Fearing defeat from outdated technology.
I told in one of the previous episodes how this old Atari console crushed ChatGPT. And now Google Gemini didn’t even lose to it. It refused to play chess with the legendary Atari console after learning about its competitors’ defeats.
Engineer Robert Caruso, who had already organized chess battles between the Atari Chess game and ChatGPT, decided to continue the experiment with Gemini. This was especially interesting since ChatGPT and Copilot are built on the common OpenAI platform, while Gemini represents a fundamentally different multimodal model from Google.
Initially, Gemini demonstrated complete confidence, stating it would almost certainly win over Atari Chess. Which, I quote, “isn’t even close to being a large language model”. Google’s AI compared itself to a modern chess engine capable of calculating millions of moves and evaluating infinite positions.
However, after learning that other advanced systems had already lost to the old console, Gemini radically changed its position. It honestly admitted to exaggerating its chess abilities and concluded that, I quote, “the most effective and reasonable solution would be to cancel the match”.
Thus the cult gaming system with a processor of only 1.19 MHz and 128 KB of RAM didn’t get a chance to battle one of the most powerful AI systems of our time.
