American defeats invincible AI system in Go game

American Kellin Pellerin beat a state-of-the-art artificial intelligence system in a game of Go by taking advantage of a previously unknown vulnerability discovered by another computer. However, in a game in which he won 14 out of 15 games, he won without direct computer assistance.

According to Yahoo.com, the win exposed a weakness in the best software for the game of Go, which is characteristic of the AB systems widely used today, including the ChatGPT chatbot.

A winning tactic was suggested by a computer program examining AB systems for their weaknesses. Kellin Pellerin then implemented the proposed plan.

Adam Gleave, CEO of FAR AI, a California-based research company, said the program they developed played more than 1 million games against KataGo, one of the best systems in the game of go, to find a weakness that a human player could exploit.  

According to Pelrin, the winning strategy is not that simple, but it's not very complicated either: the average person can learn it and apply it to a game against a computer. He also used this method and beat another leading system in the game of go, Leela Zero.

The AlphaGo system, developed by Google research company DeepMind, defeated the South Korean world champion of the game, Lee Sedol, 4-1 in 2016. At the time, it was stated that this system could not be beaten.