Artificial Intelligence

Atari 2600 Chess Game From 1977 Beats ChatGPT

JUN 11 2025·Updated JUN 12 2025

Story

Above: Atari 2600 VCS console. Image credit: Robee Shepherd/Contributor/Moment via Getty Images

The Facts

Citrix engineer Robert Caruso has conducted an experiment pitting ChatGPT against the 1977 Atari Chess game running on an emulated Atari 2600 console. The OpenAI chatbot lost decisively on the beginner difficulty level despite the Atari console's limited 1.19 MHz processor.
TechSpot
TweakTown
The experiment began after ChatGPT volunteered to play Atari Chess during a conversation about AI and chess history, expressing confidence it could quickly defeat a system that thinks only one to two moves ahead on antiquated hardware.
Futurism
TechSpot
However, ChatGPT struggled with basic chess mechanics during the 90-minute match, confusing rooks for bishops, missing obvious pawn forks, and repeatedly losing track of piece positions on the board. It required constant correction from Caruso to prevent catastrophic moves.
LinkedIn
Futurism
The AI initially blamed its poor performance on the abstract visual symbols used by the Atari 2600's low-resolution graphics to represent chess pieces but continued making critical errors even after Caruso switched to the standard chess notation format.
Retro Dodo
Tom's Hardware
Despite ChatGPT's access to modern GPU infrastructure and language processing capabilities, the simple 8-bit chess engine from 1977 consistently outperformed the chatbot using basic brute-force board evaluation without any advanced algorithms or machine learning.
TweakTown
The Register

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-skeptic narrative

This experiment perfectly exposes the fundamental flaws of so-called "artificial intelligence." Large language models, such as ChatGPT, are essentially glorified autocomplete systems that generate plausible-sounding text without any genuine understanding, actual reasoning, or game-playing intelligence in specialized domains. The fact that a 1977 chess program can outperform ChatGPT suggests these systems are merely expensive marketing hype designed to inflate tech company valuations.

Retro Dodo

TechSpot

by Jennifer Shahade

Techno-optimist narrative

While this chess experiment is amusing, it doesn't diminish ChatGPT's remarkable capabilities in language processing, creative writing, and problem-solving across countless domains. Chess requires specialized algorithms and deep game tree analysis, which isn't what large language models were designed for. The comparison is like criticizing a race car for being bad at flying — different tools serve different purposes in the AI toolkit.

The Register

Futurism