Human brain has lost to artificial intelligence in two of the five-game challenge of Weiqi being played between Google's computer program AlphaGo and Korean Lee Sedol, who is the world best at the ancient Chinese game. Saturday's game will now be crucial.
AlphaGo in fact seems to have learnt from mistakes of the first game while beating its human competitor soundly in the second game. Onlookers were surprised by the program's unusual moves and apparent learning from the earlier game.
In a press meet after the second game, Lee Sedol was all praise for AlphaGo, saying, "Yesterday I was surprised, but today, more than that, I'm quite speechless." To a question on his defeat, he said it was because he couldn't find any weaknesses in his opponent.
The challenge began on March 9 and will end on 15th. It was last October that AlphaGo beat Europe's human champion, Fan Hui. Sedol holds the world crown.
According to Google, the game also known as Go has more potential moves than atoms in the universe, making it very complex. No wonder, there is a $1 million to be won in prize money by the winner. If AI wins, Google plans to donate the money to charities.
Like chess, the game involves two players placing black or white stones on a huge square board, while trying to capture their opponents' stones or secure empty spaces on the board. But unlike a game of chess there are ten times more possible moves from one position, calling for human-like intuition often.
AlphaGo is the brainchild of Google's DeepMind.
Computer speed and memory capacity doubles every few years but games are where a machine can exhibit human intellect. AlphaGo is believed to be superior to IBM's Deep Blue that defeated Kasparov in 19197 and supercomputer Watson that beat humans in a quiz show. AlphaGo picks up tips on the go, from game to game, through trial and error.
"AlphaGo is really more interesting than either Deep Blue or Watson, because the algorithms it uses are potentially more general-purpose," Nick Bostrom of Oxford University's Future of Humanity Institute told AFP.
Creating multi-purpose intelligence, built on many inputs and self-learning like humans do is the ultimate goal in AI. Such a "human-like" approach is what AlphaGo works with using two sets of "deep neural networks" containing millions of connections similar to neurons in the brain.
This raises the spectre of a world run over by hostile AI for some like Stephen Hawking and Elon Musk, but for experts it is about putting values into machines and getting them to render useful tasks. Today, there are already many million robots employed in factories, hospitals, homes and warehouses.
"The winner here, no matter what happens, is humanity," Eric Schmidt, executive chairman of Google's parent company Alphabet, told at a press conference in Seoul at the opening of the tournament on Tuesday.
"Humanity wins because the advances in artificial intelligence and machine learning will make each, and every other, human being in the entire world smarter, more capable - just better human beings."
But not all think alike. In their open letter issued last year, Hawking et al have asked scientists to look at the opportunities and risks associated with increasingly intelligent machines. Cheap to mass produce and ideal for use in assassinations, destabilising nations, or selectively targeting ethnic groups, AI comes with much danger. Oxford University lists AI among top 12 threats to humanity. Least of the dangers is a loss of jobs to the machines as they learn to predict weather and financial markets, drive cars and recognise nuances of language!
The Go games can be followed here.
