Chap claims Atari 2600 'absolutely wrecked' ChatGPT at chess
1.19MHz eight-bit CPU trounced modern GPUs – can you do better with your retro-tech?
The Atari 2600 gaming console came into the world in 1977 powered by an eight-bit processor that ran at 1.19MhZ, and just 128 bytes of RAM – but that’s apparently enough power to beat ChatGPT at chess.
So says infrastructure architect Robert Caruso, who over the weekend posted the results of an experiment he conducted to “pit ChatGPT against the Atari 2600’s chess engine (via Stella emulator) and see what happens.”
ChatGPT confused rooks for bishops, and repeatedly lost track of pieces
Caruso decided to run the experiment after conversing with ChatGPT about the history of chess. During that chat, Caruso says ChatGPT volunteered to play against "Video Chess" – the game Atari commissioned to run on the 2600.
Video Chess debuted in 1979 and in the years since chess wonks have discussed the merits of the game. Our reading of their opinions suggests the game plays at a level beginners may find challenging, and might give regular recreational players of intermediate skill a little to worry about.
Caruso thought his experiment would be “a lighthearted stroll down retro memory lane.”
Instead, he watched as the Atari humiliated ChatGPT.
“ChatGPT got absolutely wrecked on the beginner level,” he wrote.
“Despite being given a baseline board layout to identify pieces, ChatGPT confused rooks for bishops, missed pawn forks, and repeatedly lost track of where pieces were.”
- Atari 400 makes a comeback in miniature form
- Christmas 1984: The last hurrah for 8-bit home computers
- Weeks with a BBC Micro? Good enough to fix a mainframe, apparently
- The New ROM Antics – building the ZX Spectrum 128
Caruso said ChatGPT explained its poor performance by stating the icons Atari chess uses are “too abstract to recognize”. But even after changing to standard chess notation, the chatbot “made enough blunders to get laughed out of a 3rd grade chess club.”
“For 90 minutes, I had to stop it [ChatGPT] from making awful moves and correct its board awareness multiple times per turn,” Caruso wrote. The chatbot “kept promising it would improve ‘if we just started over’.”
Eventually the bot conceded the game.
“Have you played Atari today?” Caruso asked, invoking the company’s advertising slogan.
“ChatGPT wishes it hadn't,” he concluded.
A challenge to readers
The Register knows many readers enjoy and operate retro-tech devices.
We challenge you to make them fight AI and let us know the results.
Send news of the results to us here. ®