Why Google’s DeepMind developed Go-playing AI – and how it may impact our lives

PUBLISHED : Wednesday, 24 May, 2017, 10:05pm
UPDATED : Wednesday, 24 May, 2017, 10:59pm

Mastering the most complex board game in human civilisation is merely the beginning of Google’s venture into artificial intelligence as the company pledges to use the technologies behind its world-class Go-playing computer programme for more general purposes, from health care to education.

“Games are the best [way] to train AlphaGo, but we don’t want to stop at games,” said Demis Hassabis, founder and CEO of Google’s AI subsidiary DeepMind, adding that he wants to apply the technologies behind the company’s Go-playing AI in more domains.

He didn’t elaborate on the company’s plans, only saying that machine learning technology and reinforcement learning technology, which were the reasons for AlphaGo to master Go, can be used to solve real world challenges, especially in drug discovery and material design. Go is an ancient board game that was believed to be impossible for a computer to crack.

Hassabis’ comments came on Wednesday in Wuzhen, eastern China’s Zhejiang province, where AlphaGo scored its first victory in a three-game match against Ke Jie, the world’s top Go player, extending the machine’s edge over humanity in a contest to redraw the boundaries between human intelligence and the artificial variety.

AlphaGo wins first match of three over Chinese Go master with ability to surprise

Unlike those who worry about a sci-fi nightmare in which robots supersede humans, Hassabis believes AI is just like the Hubble telescope – “a scientific tool that allows people to see farther and better understand the universe”.

The union of humans and AI-powered machines will unlock the potential of mankind and bring inspirational ideas, like how AlphaGo inspired professional Go players to come up with new tactics and strategies, he added.

AlphaGo first entered the limelight in March 2016 when it surprisingly beat South Korean grand master Lee Sedol, the then reigning world champion, in a 4-1 match.

Computers have trumped humans in several board games since IBM’s DeepBlue defeated the chess master Gary Kasparov in 1997. But AlphaGo’s victory is seen as the landmark event for the development of AI as the game requires decisions to be made on intuition rather than calculation. Put another way, AlphaGo must adapt to the gaming situation in front of it and no pre-programming can help it win.

Many of the moves made by AlphaGo in its first game against Ke Jie on Tuesday were described by commentators as “unexpected”.

“There was a cut that quite shocked me, because it was a move that would never happen in a human-to-human Go match,” said Ke Jie in a post game press conference on Tuesday, adding that there are lots of things worthwhile learning and exploring from AlphaGo.

Researchers at Google now envision an AI-enabled future in which Chinese doctors can ask their AI assistants to search all the research papers on certain diseases online, and not only translate them into Chinese, but analyse the material and provide only the essential information.

With the huge amount of data involved in health-related industries, Eric Schmidt, executive chairman of Google’s parent company Alphabet, said health care will be impacted the most by the application of AI in the coming five to 10 years.

“If I were at the age of 22, I’d be joining companies like DeepMind,” Schmidt said in Wuzhen on Wednesday. The quickest way to make money is to apply AI technologies in health care and businesses that boast a large amount of data, he added.

An increasing number of start-ups have been entering the health care AI space, focusing on areas from virtual nursing to drug discovery. Research from CB Insights shows that deals involving health care-focused AI start-ups grew from less than 20 in 2012 to nearly 70 in 2016.

2016 saw two new unicorns – private companies worth US$ 1 billion or more – emerge in the AI-enabled health care space: China-based health care big data platform iCarbonX and oncology-focused Flatiron Health.

“By 2025, AI systems could be involved in everything from population health management to digital avatars capable of answering specific patient queries,” said Harpreet Singh Buttar, an analyst at Frost & Sullivan.