DeepMind’s Demis Hassabis is AI’s grandmaster
Growing up in north London, the child of a Greek Cypriot father and a Chinese Singaporean mother, Hassabis was a child prodigy in chess from the age of 4. He began writing his own computer games at 8, created one of the first video games to use AI at 17, and founded his own video game company not long after graduating from Cambridge University at 20.
So perhaps it makes sense that Hassabis’s AI startup DeepMind, founded in 2010 and sold to Google just four years later, would achieve its first major successes with AI models that used deep reinforcement learning to rapidly master video games like Space Invaders and Q*bert without any knowledge of the actual rules.
That was followed with AlphaGo, which learned the ancient strategy board game of Go and would in 2017 defeat the world’s number one human player — an event that did perhaps more than anything else to awaken the world to the rapid progress of AI. New models could dominate a variety of games even faster, reducing the time and human intervention needed to acquire mastery.
Games are a logical playground for both AI models and for the men and women who design them. Games have clear rules and clear metrics for success and failure. When IBM’s Deep Blue defeated world chess champion Garry Kasparov in 1997, it was considered a major landmark in the advancement of AI. But whereas Deep Blue triumphed primarily thanks to sheer computational force, which enabled it to examine 200 million moves per second, the models Hassabis helped shepherd at DeepMind seemed capable of truly learning, at least within the bounds of the games.
But as useful as games are as a testing ground for AI capabilities, they’re limited in their application and usefulness in the far more messy real world. (An AI that can beat any human player in the strategy game Starcraft II, as DeepMind’s AlphaStar could, is neat, but won’t exactly change the world.)
That’s what makes DeepMind and Hassabis’s more recent turn to exploring the value that advanced AI could bring to biology research all the more important.
In November 2020, DeepMind’s AlphaFold model achieved world-beating results in one of the grand challenges of life sciences: correctly predicting the three-dimensional structure of proteins from their one-dimensional amino acids sequences. By July of this year, AlphaFold could predict the structure of some 200 million proteins from 1 million species, covering just about every protein known to human beings. It could do in as little as 10 to 20 seconds what might have taken human researchers weeks or more of work in the lab.
Hassabis has called AlphaFold “a lighthouse project, our first major investment in terms of people and resources into a fundamental, very important, real-world scientific problem.” He’s right. The recent progress in AI has been astounding, but we have yet to see the application that could do truly meaningful work faster and better than human beings can. AlphaFold, which promises to put a jetpack on some of the most vital work in biology and speed the pace of desperately needed research, promises to be just that. Hassabis clearly agrees — he now splits his time between DeepMind and Isomorphic Labs, an Alphabet startup that will use AI to design drugs.
Source: Vox.com