When it comes to games that use strategy to win, it seems impossible to beat a machine since it is already programmed to know all the possible steps and how to counter them. However, a human was able to find a weakness in the system and beat the machine in a game of Go.
Human Player vs. AI
Kellin Pelrine, the human factor in the game, defeated the top-ranked AI system at Go, a strategy board game whose aim is to gain more territory than the other. Pelrine is an American player who placed second in the top amateur ranking for Go.
Beating strategy with research, a flaw was found by another computer and Pelrine exploited that flaw which won him the game. He won 14 out of 15 games in a head-to-head confrontation without computer support, according to Ars Technica.
This revealed a weakness in the Go computer programs that are also used by most of the present AI systems, which include the famous ChatGPT. The tactics used to win were said to be suggested by a computer program that was looking for weaknesses in the AI system.
Adam Gleave, the CEO of FAR AI where Pelrine works as a machine learning researcher, and the research company that designed the program, said that it was surprisingly easy for them to exploit the system.
The software, in order to learn, played one million games of KataGo, one of the top Gopplaying systems, which they used to find a blind spot that a human player could exploit. The strategy they found was "not completely trivial" but it's also not very difficult for a human to learn.
Perline claims that the flaw can be used by an intermediate-level player to beat their AI opponent. He also used this exploit to win against another Go system called Leela Zero, marking a victory for humans against AI, even if it is with the help of a computer system.
Humans Surpassing AI
There have been some concerns since AI has become more and more prominent along with technology. Although it can do things more efficiently than humans, it can never replace humans entirely without losing a few important factors.
At most, artificial intelligence is there to assist and not to replace. It's safe to say that our AI technology is still in its early stages, and it's still pretty dumb. It cannot answer complex questions that still need human emotion and reasoning.
Machines are good at patterns and algorithms, and they can only match patterns they have learned, as mentioned by Diginomica. This shows AI's limited capacity as opposed to humans who can learn unlimited patterns, brought on by millions of years of evolution.
It hasn't even reached advancement to the point that it can take spoken food orders from humans, which can be seen with McDonald's AI-powered drive-thru machine. It cannot take simple orders much less the ones with certain modifications.
