We've been told repeatedly by boorish tech companies that AI is soon going to overtake humanity when it comes to intelligence, but it would seem that playing chess is not a strong suit of these billion-dollar "Large Language Models".
As reported by Extreme Tech, Citrix Engineer Robert Caruso recently pitted OpenAI's ChatGPT 4o model against Atari's 1979 Video Chess cartridge on the Atari 2600, and even on the easiest setting, ChatGPT got its ass handed to it on a plate.
"ChatGPT got absolutely wrecked on the beginner level," said Caruso. "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—first blaming the Atari icons as too abstract to recognize, then faring no better even after switching to standard chess notation."
The match lasted an hour and a half, and at one point, Caruso even stepped in to assist ChatGPT because it was doing so badly. "Meanwhile, Atari’s humble 8-bit engine just did its thing," adds Caruso. "No language model. No flash. Just brute-force board evaluation and 1977 stubbornness. For 90 minutes, I had to stop it from making awful moves and correct its board awareness multiple times per turn. It kept promising it would improve “if we just started over.” Eventually, even ChatGPT knew it was beat — and conceded with its head hung low."
Of course, AI evangelists, inbetween telling us all of our jobs are at risk, will be quick to point out that ChatGPT is not a chess program. That's a fair comment, but until it can beat a near-50-year-old chess video game, I'm not going to worry about The Terminator becoming a reality just yet.