Vintage Atari Console Outsmarts Modern AI In Chess Game
Technology is a constantly evolving force a modern marvel to many yet a source of apprehension for others who see it on the verge of radically changing society. The mere mention of advanced technology can be startling.
If you asked someone to pinpoint the most advanced technology of today many would likely name ChatGPT the super brainy chatbot from OpenAI. This AI is so packed with artificial intelligence it often makes conventional intellect seem outdated. But if ChatGPT represents the peak of smart technology how did it recently lose to a video game console released in 1977.
The Unlikely Showdown Atari 2600 vs ChatGPT
When considering competitors for ChatGPT names like Google Gemini Microsoft Copilot Meta AI Anthropics Claude Perplexity AI and DeepSeek usually come to mind. The Atari 2600 a 48 year old home video game console famous for bringing PacMan to countless homes is not a name you would expect in this lineup.
However Citrix Engineer Robert Jr Caruso changed that perception. He used the retro console from Atari to repeatedly defeat the GPT4o model from OpenAI in a simple game of chess. In a now viral LinkedIn post Caruso described his 90 minute experiment. He pitted the computing power of thousands of Nvidia GPUs against the Atari 2600s single 1.19 MHz 8 bit MOS Technology 6507 processor. Caruso stated that ChatGPT made enough blunders to get laughed out of a 3rd grade chess club.
Understanding The Victory A Matter Of Specialization
This victory of old tech over new is a classic David versus Goliath story and a surprising result for many. It seems they do not make them like they used to.
However there is a vast difference between a general purpose chatbot and a dedicated chess engine. The Atari 2600 in this experiment ran via the Stella emulator not as physical hardware. This emulated system likely had an advantage from the beginning. While OpenAI has significantly improved the memory and reasoning capabilities of its models ChatGPT is still fundamentally a language prediction machine not a specialized chess grandmaster like Deep Blue.
The chess program for the Atari 2600 could only predict two moves ahead. Yet its ability to defeat ChatGPT highlights current shortcomings of the AI in specific strategic tasks. This outcome pokes a bit of fun at the notion of ChatGPT being the most intelligent chatbot in the world a claim OpenAI itself has rarely made.
OpenAIs Perspective On AI Development
In fact OpenAI has often been modest about the current capabilities of its AI. As far back as 2023 in an episode of the Lex Friedman podcast OpenAI CEO Sam Altman described GPT4 as a very early AI. He noted it is slow it is buggy and it does not do a lot of things very well. Later in 2024 during a Q&A at Stanford University Altman remarked that ChatGPT was running on the dumbest model any of you will ever have to use again by a lot.
Only recently has Altman suggested that AI is beginning to meet its high expectations. In a blog post published recently titled The Gentle Singularity Altman predicted We do not know how far beyond human level intelligence we can go but we are about to find out.
The Future Of AI Is Still Unfolding
So while the classic console from Atari may have outplayed ChatGPT in this instance this victory for retro technology could be very brief. If words from Sam Altman are any indication the capabilities of AI are set to expand dramatically and rapidly. The game of chess between old and new tech serves as a fascinating snapshot of the current stage of development for AI.
