Beating the top human Go player in the world is yet another example of how artificial intelligence (AI) is breaking new ground. But in March 2016 that’s exactly what happened. Google’s DeepMind AlphaGo AI trounced legendary Go champion Lee Sedol 4-1.
It’s not the first AI to beat a human at its own games. IBM’s Deep Blue beat Chess Grandmaster Garry Kasparov in 1997, while its current Watson supercomputer AI beat human opponents at the game show Jeopardy in 2011. In 2014, Google let its DeepMind software learn and play Atari video games like Breakout and Time Pilot, just from the raw pixel inputs.
Go was undoubtedly a grander challenge. There are billions upon billions of possible board configurations in the 2,500 year-old board game. “More than the number of atoms in the universe,” wrote Demis Hassabis on the Google Blog, who added that the challenge for an AI was “more than a googol times larger than chess.”
So what’s next? In terms of gaming, there’s already been talk of AlphaGo tackling Blizzard’s StarCraft 2, the complex real time sci-fi strategy game. But playing games is only a demonstration of intent.
“The aim of DeepMind is not just to beat games, fun and exciting though that is,” Hassabis told The Verge. “They’re useful as a testbed, a platform for trying to write our algorithmic ideas and testing out how far they scale and how well they do and it’s just a very efficient way of doing that. Ultimately we want to apply this to big real-world problems.”
What sort of problems? AI could target bigger challenges across finance, travel, robotics, healthcare and even the arts. In some cases, this has already begun.
In January, the BBC reported that the hedge fund Aidyia had begun using an algorithm to trade on the US stock market. The company uses “cutting edge artificial general intelligence (AGI) technology to identify patterns and predict price movements,” allowing it to respond quicker, identify trends faster and, ultimately, make more money for its customers.
Also in the US, Google’s self-driving car system could become the first non-human to qualify as a driver. In a letter to Google from the US National Highway Traffic Safety Administration, the NHTSA agreed that a Google SDV “will not have a ‘driver’ in the traditional sense” and that it will interpret ‘driver’ as “referring to the (self-driving system), and not to any of the vehicle occupants.”
If the NHTSA sticks to this line of thinking, it paves the way for officially sanctioned autonomous vehicles, arguably the biggest change to transportation in over 100 years.
Not only can AIs drive, they can apparently write. The Sky Arts TV series ‘Computer Says Show’ follows an experiment to see whether a computer (and various software tools) can assist in the composition of a West End musical.
Working with statistical data that encompassed all aspects of hit musicals (including cast size, emotional structure and popular songs), Dr James Robert Lloyd, Dr Alex Davies and Professor Sir David Spiegelhalter from Cambridge University used computer algorithms to construct their own show. The result, Beyond The Fence, tells the story of a mother and daughter at Greenham Common peace camp in 1982.
In the same vein, computers are also writing books. The LA Times reports that four entries to the Nikkei Hoshi Shinichi Literary Award in Japan were co-authored by algorithmic authors. “Humans decided the plot and character details of the novel,” says the paper, “then entered words and phrases from an existing novel into a computer, which was able to construct a new book using that information.”
If all of these examples continue to develop, our children will be growing up in a world where their money is invested by computer, their cars are driven by them, and the content they watch, read or listen to is algorithmically authored, even personalised for them. Where will us humans fit into a world run by the descendants of AlphaGo? Perhaps while computers will play by the rules, it will be our job to break them.
Find more gaming resources from Intel here