AlphaGo
Matrix rising? BY HENRY THOMPSON
Hollywood’ s ultimate nightmare of a takeover by the machines has come a little closer after a super-computer beat a master of the 2,500-year-old board game Go
Science-fiction has long dramatised the emergence of artificial intelligence able to out-think the whole of humanity. Think the Terminator or the Matrix which posited the rise of sentient machines, and you have some idea of the astonishing implications of Google’ s Deep Mind super-computer called AlphaGo.
In March, the Deep Mind machine beat the Korean grandmaster and European champion of the strategy game by four games to one. But what shocked watching experts in the board game and computer algorithms was that AlphaGo played in non-obvious ways, making sly, even bizarre moves.
“ The really significant thing about AlphaGo is that it( and its creators) cannot explain its moves,” John Naughton reported in his Observer Networker column.
“ And yet it plays a very difficult game expertly. It’ s displaying a capability eerily similar to what we call intuition – knowledge obtained without conscious reasoning.”
What the experts were saying is that AlphaGo is not just a number-cruncher using its massive computer memory as IBM’ s Deep Blue did when it beat chess grandmaster Gary Kasparov back in 1997.
“ Not so: the number of possible positions in Go outnumber the number of atoms in the universe and far exceed the number of possibilities in chess,” added Naughton.
“ If AlphaGo really demonstrates that machine could be intuitive, then we have definitely crossed a Rubicon of some kind.”
The traditional Chinese game of Go involves two players taking turns to place black and white marker discs known as stones on a 19 squares by 19 squares grid. Players win by taking control of the most territory on the board, by surrounding opponent’ s stones with their own.
AlphaGo was developed in London by Google’ s DeepMind UK branch, and specifically designed to challenge grandmasters of Go. No wonder the words of Sir Alex Ferguson resonate with millions of gamers, all around the world:“ As long as there are games to play, then it is not over.”
One man in particular, Lee Se-dol, from South Korea, would have sympathised, faced with his toughest opponent to date. He knew AlphaGo had already beaten a professional- in October 2015, defeating the European Go champion Fan Hui.
That was the first time a computer programme had beaten a professional human player on a full-sized board without a handicap.
When Lee Se-dol faced AlphaGo, he had the second highest number of Go international championship victories in the world. According to the Korean Herald, Mr Se-dol was considered the fourth greatest Go player in the world currently.
The first three-of-five games ended in victory for AlphaGo. But Mr Se-dol beat AlphaGo in the fourth game, when the programme resigned at the 180th move. AlphaGo won the fifth game on Mr Se-Dol’ s resignation.
Shortly after the second game, Mr Se-dol said that he felt“ speechless. From the very beginning of the match, I could never manage an upper hand for one single move. It was AlphaGo’ s complete victory.”
Lee said his inevitable loss to the machine was saddening, but that“ robots will never understand the beauty of the game the same way that we humans do”.
Lee went on further to say that his game 4 win was“ a priceless win that I would not exchange for anything”.
Most experts were expecting the Korean to beat the AlphaGo programme easily. Murray Campbell, creator of the Deep Blue software, described AlphaGo’ s win as“ the end of an era … board games are more or less done and it’ s time to move on”.
AlphaGo won a prize of $ 1 million, with the cash being donated to several charities, primarily UNICEF. Mr Se-dol received $ 150,000 for participating and an additional $ 20,000 for his win.
As Adam Roberts wrote in the Guardian:“ Being able to crunch huge numbers really quickly is not the same thing as intelligence, and certainly not the same thing as sentient self-consciousness.
“ The ability to intuit, to make leaps of comprehension – not just to extrapolate according to pre-programmed rules but to speculate – is a lot closer to the Holy Grail of proper AI.
“ AlphaGo’ s achievement is more modest.”
25