When DeepMind’s AlphaGo program beat the world champion in the game of Go back in 2016, I instantly assumed that it must have taken DeepMind’s computer scientists decades to build a machine capable of playing and beating humans in such a game more complex than chess.
Well, today I learned that the program learned to play the game at a world-champion level in a matter of hours using reinforcement learning. In the TED Tech podcast episode How AI is unlocking the secrets of nature and the universe, DeepMind’s CEO Demis Hassabis discusses the motivations to build such a system and how it could help scientists decode the mysteries plaguing humans for thousands of years.
Hassabis talked about protein folding, the benefits of understanding it (mainly fighting incurable diseases more efficiently), and why it is so difficult to study. He said there are more than 200 million varieties of proteins and it takes years to fully understand just one of them. AlphaFold, an AlphaGo successor, recently decoded them in a matter of days, saving the scientific community billions of years of valuable time.
That’s the real promise of artificial intelligence. Talk about AI beyond ChatGPT 😉