Life Tech Google AI beats human 4-1
Updated:

Google AI beats human 4-1

lee se-dol
Getty
Share
Twitter Facebook Reddit Pinterest Email

Google’s artificial intelligence (AI) program AlphaGo has won the final game of its five-match challenge against Go grandmaster Lee Se-dol, taking the series 4-1.

AlphaGo returned to its best form for the finale, using the much-hyped “intuitive” AI of its formidable neural networks to outwit Mr Lee.

Before the match, AlphaGo was given an honorary “ninth dan” professional ranking, equivalent to that held by Mr Lee, who has 18 international titles to his name and is widely considered one of the greatest Go players of the modern era.

Google’s stunning opening match against Go champ
A surprising rival to Apple, Samsung emerges
Earth sets terrifying temperature record

The rank is reserved for those whose ability at the ancient board game borders on “divinity”.

“Neither man nor machine has ever received a honorary professional dan certificate to date, although many people have received honorary amateur certificates”, an association spokeswoman said.

The human champ did manage to win one match. Photo: Getty
The human champ did manage to win one match against the computer. Photo: Getty

The citation on the certificate, serial-numbered 001, said it was given in recognition of AlphaGo’s “sincere efforts” to master Go’s Taoist foundations and reach a level “close to the territory of divinity”.

The AI won the first three games against Mr Lee fairly easily, but the 33-year-old South Korean then fought back to take the fourth match.

The final match was a gruelling five-hour affair, with AlphaGo eventually outsmarting Mr Lee.

The South Korean held his head in his hands at the table after he resigned

For AlphaGo’s creators, victory goes far beyond the $US1 million-dollar prize on offer in Seoul, proving that AI can go beyond superhuman number-crunching.

Go involves two players alternately laying black and white stones on a chequerboard-like grid of 19 lines by 19 lines.

The winner is the player who manages to seal off more territory. AlphaGo uses two sets of “deep neural networks” that allow it to crunch data in a more human-like fashion — dumping millions of potential moves that human players would instinctively know were pointless.

It also employs algorithms that allow it to learn and improve from experience.

Comments
View Comments