Modern LLM vs 1970s Atari: Gemini admits it can't win, walks away
Google's Gemini chatbot reportedly refused to play a game of Chess with the Atari 2600 after it learned that the vintage game console had already taken down ChatGPT and Microsoft Copilot.
Infrastructure Architect Robert Caruso told The Register that the question intrigued him because "while ChatGPT and Copilot are cousins built on the same OpenAI base, Gemini is a completely different beast."
Caruso had previously pitted the Atari 2600 with its paltry 1.19 MHz 8-bit processor and a whopping 128 bytes of RAM against the might of ChatGPT for some very interesting results.
After wrecking ChatGPT, the Atari 2600 then took on Microsoft's Copilot, and the results were much the same. Interestingly, in both cases, the AI chatbots seemed to have a misplaced sense of confidence and boasted about their prowess in Chess.
Google's Gemini did this as well. However, when Caruso informed the chatbot of Atari 2600's previous bouts, it seemed to backtrack and admit it had "hallucinated its Chess prowess," showing a rare sense of internal skill assessment for an AI.
Gemini then decided that "canceling the match is likely the most time-efficient and sensible decision." Caruso said that he was impressed by Gemini's ability to identify its shortcomings.
"Adding these reality checks isn't just about avoiding amusing chess blunders. It's about making AI more reliable, trustworthy, and safe - especially in critical places where mistakes can have real consequences," Caruso told The Register.
Source(s)
The Register | Robert Jr. Caruso on LinkedIn
Image Source: Evan-Amos on Wikimedia Commons
