In 1955 John McCarthy, an assistant professor at Dartmouth College in New Hampshire, organised a group “To clarify and develop ideas about ‘Thinking Machines’.” The following year, he coined the term “Artificial Intelligence”, now shortened to AI.
The early researchers soon realised that the world of board games represented a perfect training ground for testing machine intelligence. Games such as Chess and GO had long been held to demonstrate the power of the human mind.
This was first proposed in Arthur Samuels’ 1959 paper “Some Studies In Machine Learning Using the Game of Checkers.” He wrote a program which was able to learn from experience and was soon able to compete against amateur players. However, it took until 1995 for the game of Draughts to be “solved”. This means that the computer is able to process all 5 X 10^ 20 possible combinations of the game, which in turn means that a human can never win. However, Draughts was not considered to be a game on the level of GO, or even Chess, where the possible combinations are much greater. In fact, in GO there are calculated to be more possibilities than there are atoms in the Universe. And so it became standard throughout the Seventies and Eighties for Chess players to declare that humans would always be supreme, unless, as one player predicted, a human would only be defeated if he (sic) were playing many games simultaneously and made a once in a lifetime error.
However, AI programmers continued to make progress, as one unfortunate human discovered on the very day he became World Backgammon Champion. He was defeated by AI on the same evening.
Google acquired a company known as “Deep Mind” and its CEO, Demis Hassabis, was soon on hunt for
Chess champions to defeat. And so it came to pass that world champion Garry Kasparov encountered Chess-playing program – “Deep Blue.” At first, all went well for humanity; Kasparov won the first encounter easily. However, new iterations of Deep Blue were produced and in May 1997 it won a series of matches 3 ½ - 2 ½. Since then the programs have continued to become more powerful and it is now accepted that humans can never defeat them. This realisation is summed up by (human) World Champion Magnus Carlsen when he revealed that he won’t play against AI opponents, saying that “He just loses all the time and there’s nothing more depressing than losing without ever being in the game.”
However, GO is a much more complex game than Chess and for many years AI performed very badly against human experts. Then the Deep Mind programmers released “Alpha Go” in 2016. It went on to defeat Lee Sedol, a 9 dan professional in a No-Handicap match in the same year. Finally human supremacy was brought to an end in 2017 when it defeated Ke Jie, who at the time was World Number One, and had been for two years. Reflecting on his defeat, he said, “ I'm a little bit sad, it's a bit of a regret because I think I played pretty well."
However, “pretty well” was not good enough because there is now no point in programs like Alpha Go playing humans. The Deep Mind team went on to release “Alpha Zero” in December 2017 which required only 24 hours of training to make play against a human pointless. Instead, it plays other programs, defeating fellow programs Stockfish and Elmo, as well as earlier versions of itself.
The obvious rejoinder to these claims of machine superiority is to say, “It’s only a game” and to dismiss such activities as trivial. However, this is not how the World Champions and philosophers saw the possibility of defeat before it happened. Most denied it could ever happen, but believed it would be a dark day if it ever did.
The fact remains that these activities were not seen as “only games” at the time but demonstrations of the power of the human mind.
And now, humans who beat their human opponents must realise there is at least one player it is mathematically impossible for them to conquer.
Comments