In the field of artificial intelligence, ChatGPT is widely popular due to its powerful language generation capabilities. However, a recent experiment has led it to an unexpected failure. Engineer Robert Caruso simulated a chess match between ChatGPT and a classic game console from 1977 - the Atari 2600 (Atari2600). The result was surprising; not only did ChatGPT lose the game, but it also made multiple mistakes.
This match used the 1979 Atari "Video Chess" game. After 90 minutes of competition, ChatGPT performed extremely poorly. Caruso shared the details of this experiment on LinkedIn, pointing out that ChatGPT mistakenly identified the "rook" as a "bishop," missed opportunities for pawn forks, and frequently got confused about the positions of the chess pieces. Despite ChatGPT's attempt to complain about the abstract icons of the Atari game, its performance remained poor even when switching to standard chess notation. Caruso joked: "The mistakes ChatGPT made are enough to make it laughed at in a third-grade chess club." At the beginner level, ChatGPT was thoroughly defeated.
It is worth noting that although computers were once considered symbols of superhuman ability in chess competitions, such as IBM's Deep Blue defeating the chess grandmaster Garry Kasparov in 1997, this experiment does not mean that ChatGPT is useless in the field of chess. After all, ChatGPT is a language model, not a supercomputer specifically designed for chess matches, so its ability in playing chess seems relatively limited. Perhaps, instead of playing against it, discussing chess with it might be more appropriate.
This experiment not only reveals the limitations of artificial intelligence in specific fields but also provokes deeper thoughts about technology. Although ChatGPT performs excellently in many areas, it still needs to work hard in complex chess matches.
