Roundup: AI not smarter than humans despite beating human Go champion

Although the Google-developed Go-playing machine AlphaGo Thursday scored a second victory against a South Korean grandmaster, experts believed artificial intelligence (AI) is not as smart as humans when it comes to tackling complex, real-world problems.

After shocking the world by defeating Lee Sedol, the winner of 18 world championships, in their first match on Wednesday, the AlphaGo computer won the second one after a four-and-a-half-hour game.

The victories in the ancient Chinese board game, recognized as one of the most creative games ever devised, put the supercomputer team one victory away from claiming the prize worth 1 million U.S. dollars.

Lee, 33 years old, was quite speechless after the second defeat. "From the beginning there was no moment I thought I was leading," he told a press conference on Thursday.

The three remaining games run till next Tuesday, and experts believed AlphaGo would emerge as the final winner of the match.

"It is much more likely that the average performance of the computer is higher than the level of the human, and therefore it will continue to win almost all of the time," said Felix Hill, AI researcher at the University of Cambridge's Computer Laboratory.

Marc Deisenroth, from the Department of Computing at Imperial College London, said AlphaGo combined Deep Learning techniques and Monte Carlo Tree Search (MCTS) so as to play the game at a professional level.

MCTS is a heuristic search strategy, he said, that analyzes the most promising moves in a game by expanding the search tree based on random sampling of the search space.

"We can compare AlphaGo to Deep Blue, the computer program that beat Gary Kasparov in chess in the 90s: It is a computer program that can solve one particular task very well -- way better than humans," he added.

If AI finally wins, the prize money will be donated to the United Nations International Children's Emergency Fund, Go organizations and charities.

However, even if the supercomputer beats the professional Go player in all five games of the match, it does not mean AI is much smarter than humans.

Hill pointed out the game of Go is massively more constrained than the real world, saying that although the game contained many billions of combinations of things that can happen, the possibilities are still finite, discrete and can be easily described.

"In real life, whether computing how things move through the air or trying to interpret or produce language, there are infinitely many possible actions at any one time, and an infinite number of times such a decision must be made," he said.

"Go is much more like learning how to multiply numbers (which computers did in the 1940s) than learning to master many real world problems including understanding language," he added.

Deisenroth shared similar views with Hill. The lecturer in statistical machine learning believed computer programs are not at the level of general intelligence that humans exhibit.

"Some features of human learning are currently difficult to achieve by AI systems, such as the general ability to transfer knowledge from one problem to a new one, the ability to learn from limited experience, and the ability to reason at abstract levels or the ability to cooperate with other humans," Deisenroth said. Endi