Previous schools of thought may have seen Artificial Intelligence (AI) as a ‘fun’ tool, which only had ‘super human’ or extraordinary capabilities in the worlds of fantasy and film. Over the years AI abilities have been growing, yet until recently were no match for specially designed programs or even humans in some cases.
These conceptions are all being challenged by one of the latest AI developments, in which AlphaZero, the game-playing AI, has beaten the world’s best chess-playing computer program, after teaching itself to play in just under four hours. AlphaZero, created by DeepMind is a repurposed game playing AI which has already been victorious against the world’s best ‘Go’ players, and also against world class ‘Shogi’ players.
This development is especially exciting as previously chess had been considered to require “too much empirical knowledge for a machine to play so well from scratch, with no human knowledge added at all”, according to former world chess champion Garry Kasparov. According to creators DeepMind, the reason behind AlphaZero’s victory is a ‘more human-like approach’ compared to its competitors, in which no prior input, aside from the basic game rules, is received. The rest is then worked out in the same way a human player would learn, through repetition, playing oneself and using reinforced knowledge to improve.
For more information on the study and capabilities of AlphaZero, read the dedicated research paper published with Cornell University Library’s arXiv.