GO: Human beats AI by exploiting vulnerability

GO: Human beats AI by exploiting vulnerability

A strong amateur player of Go has defeated a system of artificial intelligence high-ranking after exploiting a weakness discovered by a second computer.

As reported by The Financial Timesby exploiting the flaw, U.S. player Kellin Pelrine decisively defeated the system. KataGowinning 14 of 15 games with no further computer assistance.

With help?

The victory is one of the maximum progress of humans against the machine after, in 2016, AlphaGo will defeat the professionals of this game by marking his leadership.

Pelrine’s victory was made possible by a research company called. FAR AIwhich developed a program to investigate the weaknesses of KataGo.

After playing more than a million games, he could find a weakness that could be exploited by a decent amateur player. “It’s not entirely trivial, but it’s not super-difficult” to learn, Pelrine said. He used the same method to beat Leela Zero, another of the best artificial intelligences of the game.

According to the computer, the best way to beat the AI was to create a large “loop” of stones to surround an opponent’s group and then distract it by making moves in other areas of the board.

Even when his group was almost surrounded, the computer missed the strategy. “As a human, it would be fairly easy to detect,” Pelrine said, as the surrounding stones stood out clearly on the board.

Systems shortcomings.

The flaw demonstrates that artificial intelligence systems really can’t “think” beyond their training.

Lightvector (the developer of KataGo) is certainly aware of the problem, which players have been exploiting for several months. In a GitHub post, he said he has been working on a fix for a variety of attack types that use the exploit.

We recommend. METADATARPP’s technology podcast. News, analysis, reviews, recommendations and everything you need to know about the technological world.


Kayleigh Williams