Chess-playing AI’S ‘humanity’ a game-changer

Google system shows signs of intuition and creativity that had been thought beyond a machine

Google Deepmind’s artificial intelligence program Alphazero is showing signs of human-like intuition and creativity after becoming expert at chess, scientists have said. While chess computers usually hold on to their pieces, Alphazero readily sacrifices them for a better position to win.

Chess-playing AI program starts to think like a human GOOGLE Deepmind’s artificial intelligence program Alphazero is showing signs of human-like intuition and creativity, in what developers have hailed as a “turning point” in history.

The computer system amazed observers last year when it mastered chess from scratch in just four hours, despite not being programmed how to win.

But now, after a year of testing and analysis by chess grandmasters, the machine has developed a unique style of play, suggesting the program is improvising like a human.

Unlike the world’s best chess machine – Stockfish – which calculates millions of possible outcomes as it plays, Alphazero learns from its past successes and failures, making its moves based on a “nebulous sense that it is all going to work out in the long run”, according to Deepmind experts.

When Alphazero was pitted against Stockfish in 1,000 games, it lost just six, winning convincingly 155 times, and drawing the remaining games.

Yet it was the way that it played that has amazed developers. While chess computers predominately like to hold on to their pieces, Alphazero readily sacrificed its soldiers for a better position in the skirmish.

Prof David Silver, who leads the reinforcement learning research group at Deepmind, told The Daily Telegraph:

“It’s got a very subtle sense of intuition that helps it balance out all the different factors. My personal belief is that we’ve seen something of a turning point where we’re starting to understand that many abilities, like intuition and creativity, that we previously thought were only human are actually accessible to machine intelligence as well. And I think that’s a really exciting moment in history.”

Alphazero started as a “tabula rasa” programmed with only the basic rules of chess and learnt to win by playing millions of games against itself in a process of trial and error known as reinforcement learning.

It is the same way the human brain learns, adjusting tactics based on a previous win or loss, which allows it to consider 60,000 positions per second, compared to the roughly 60million of Stockfish.

Within just a few hours, the program had independently discovered and played common human openings and strategies before moving on to develop its own ideas, such as swarming around the opponent’s king and placing much less value on individual pieces.

Matthew Sadler, a grandmaster, and Natasha Regan, a women’s international master, say it is unlike any traditional chess engine. “It’s like discovering the secret notebooks of some great player from the past,” said Mr Sadler. Garry Kasparov, the former world champion, who lost to chess machine Deep Blue in 1997, said: “Alphazero generates its own knowledge. It plays with a very dynamic style, like my own. The implications go far beyond my beloved chessboard.”

The analysis was published yesterday in the journal Science, and the Deepmind team now hope to use their system to help solve real world problems, such as why proteins become misfolded in diseases such as Parkinson’s and Alzheimer’s.

The results show that it may find things that humans would miss.