A Man Defeated The “Invincible” AI System In The Board Game Go
Kellyn Pelrine, a US amateur player ranked among the top non-professionals, defeated artificial intelligence by exploiting a previously unknown flaw found by another computer. However, in a tense match, he won 14 out of 15 games on his own, without the direct help of computer technology. This revenge points to a flaw in most current AI models, including Go or ChatGPT.
This victory exposed the weakness of the best computer programs in Go, which is characteristic of most of today’s widely used artificial intelligence systems, including the ChatGPT chatbot.
The winning tactic was suggested by a computer program that examined AI systems for weaknesses. The proposed plan was then used by Kellin Pelrin.
The winning strategy is “not exactly trivial, but not too hard” to learn by a human, and can be used by an intermediate player to defeat computers. He also used this method to win against another leading Go system, Leela Zero.
The discovery of a vulnerability in the most advanced Go machines points to a fundamental flaw in deep learning systems, says Stuart Russell, professor of computer science at the University of California, Berkeley. These systems are only able to recognize specific situations that they have encountered in the past, and cannot generalize knowledge in the way that people easily do. “This shows once again that we are too quick to attribute superhuman levels of intelligence to machines,” he said.
