Thursday, May 7, 2015

Robots, start your engines

The Indianapolis 500 automobile race has been a Memorial Day tradition in the United States since 1911, going back almost as far as the last time the Chicago Cubs won the World Series in 1908. (We Cub fans live on in everlasting hope and frustration!) The dramatic starting command of the race is: "Gentlemen, start your engines," which is modified to "Ladies and Gentlemen, ..." when female drivers are in the race.

Could this starting command be further modified for races in the decades to come?

In early 2014, Google acquired DeepMind for a reported $625 million. DeepMind is an artificial intelligence company that builds learning algorithms for applications such as recommendation systems for e-commerce. The company was founded by neuroscientist Demis Hassabis (former chess child prodigy and master gamer), Jaan Tallin (Skype and Kazaa developer), and Shane Legg (researcher).

Google's acqui-hiring of Deepmind helps it compete against other players focusing on deep learning. Facebook has recruited Yann LeCunn (former NYU professor) to head the company's artificial intelligence lab. IBM is investing $1 billion in its Watson supercomputer division that is working on deep learning to support applications such as medical diagnosis. Yahoo acquired the LookFlow team to lead its deep learning initiative. And on it goes.

The field of artificial intelligence (AI) has undergone several cycles of boom and bust since AI was christened and sent out the door with research momentum at The Dartmouth Conference of 1958 organized by Marvin Minsky and others. Periods of buoyant optimism for the technology have given
way to AI winters. But in 2014, an AI spring had returned. Google’s purchase of DeepMind  reflected the new-found confidence in what AI could accomplish in multiple business sectors of interest to Google.

In March 2015, Google's DeepMind team revealed an algorithm that can teach itself from scratch to play early computer games with skill equal to or better than a human. The team eventually plans to work on three-dimensional games. According to Dennis Hassibis: "if this algorithm can race a car in a racing game, then with a few extra tweaks it should be able to drive a [real] racing car,"

Am wondering if DeepMind could start working on a pitching algorithm for the Chicago Cubs. Help us, Dennis Hassibis, you're our only hope.