Yesterday’s defeat of legendary Go player Lee Se-dol marked yet another benchmark in the progress of artificial intelligence. The win, in South Korean capital Seoul, takes the Silicon Valley firm’s AlphaGo program 2-0 up in a five game series – and left 18-time champion Lee “speechless.”
It is the first time an AI has beaten a human at the 2,500-year-old Chinese game, using an advanced system of machine learning and neural networks. And, as the following list shows, it represents one of the final hurdles in machine’s recent gaming battles with man. Something for Stephen Hawking, a recent warner of artificial perfidy, to mull over, perhaps.
In probably the most famous example of man vs machine gaming, IBM’s Deep Blue took on chess legend Garry Kasparov in a 1996, six-game series. So confident was the Russian, that he scoffed at the chance to split the $500,000 winning purse 60/40, preferring to go all-or-nothing.
Kasparov would win his cash, taking the series 3-1 with two stalemates. But the fact he was beaten in game one by the computer shocked him, and the world. It was Deep Blue’s sacrifice of a pawn early on in that round that flummoxed Kasparov, subjecting him to “a wonderful and extremely human move,” he said afterwards. “I could feel – I could smell – a new kind of intelligence across the table.”
Deep Blue did not have the longevity of AlphaGo, which Google has developed to show its pioneering of machine learning: it was built purely to play chess, and was dismantled soon after the showdown with Kasparov. But the match had confirmed that computers were now more than capable of taking on the world’s best human minds, a milestone that has pushed AI into hitherto-unthought-of spheres.
Before there was Deep Blue, there was Maven, developed in 1986 by programmer Brian Sheppard. Its first incarnation used a set of 100 patterns to evaluate the value of letter racks. So good is Maven at picking incredible words, that it has become the official AI player of Hasbro, Scrabble’s creator.
By beating humans roughly 2/3 of the time, Sheppard’s creation has earned its title as an intelligence beyond humans at the word game. Sheppard, who holds a BA in mathematics from Harvard and PhD from Maastricht University, went on to write a highly-acclaimed thesis on the perfect playing of chess, in 2002.
You’d think a computer could boast the perfect poker face, right? Perhaps not. The card game is one realm in which AI has struggled against its human opponents – not least because poker relies on instinct, bluffing and playing patterns that are often illogical and unpredictable.
Just ask Claudico, a program developed by bods at Carnegie Mellon University in Pittsburgh, which took on four humans at a Texas Hold’em tournament in the city last May. Of the 80,000 hands played and whopping $170 million bet, $732,713 was won by the human players – all of whom ranked in the world’s top ten (the ‘actual’ figure on offer was $100,000 – the higher stakes were just part of the game).
“We know theoretically that artificial intelligence is going to overtake us one day,” said 22-year-old pro Bjorn Li after. “But at the end of the day, the most important thing is that the humans remain on top for now.”
Perhaps not, suggested Thomas Sandholm, who led Claudico’s development, adding that the game had been drawn: “It would have been no shame for Claudico to lose to a set of such talented pros, so even pulling off a statistical tie with them is a tremendous achievement.”
Sour grapes? More about guts, said Doug Polk, who took second prize: “Betting $19,000 to win a $700 pot just isn’t something that a person would do.” Polaris, another poker program, developed at the University of Alberta, has been underway for 16 years and has won the occasional game against humans. It seems the writing is on the wall.
The Stock Market
Let’s say that the markets are one giant game (a statement with no little weight), and it’s one area that computers have more than excelled at in recent history. No longer are stock market floors the mad, human stramashes of old, computers having cut out latency and scoured algorithms to buy and sell at frantic speeds.
If that sounds scary, well, it can be. In 2010 the Dow dropped by almost 1,000 points when a program from Kansas triggered an avalanche of activity from other companies’ computers. Scott Patterson, author of 2010 book Quants (a term for the quantitative analysts that monitor stocks), told Wired in 2011 that AI is “a very powerful way of investing.” But, he added, they could easily cause cycles of booms and busts. “I don’t want to demonize it,” he added. “I think there has to be a happy medium. But I’m personally worried that it can run off the rails.”