When we started DeepMind in 2010, there was far less interest in the field of AI than there is today. We wanted to try a new interdisciplinary approach to accelerate the field ? bringing together new ideas and advances in machine learning, neuroscience, engineering, mathematics, simulation and computing infrastructure - as well as a new way of organising scientific endeavour.
We saw early success in computer games, which researchers often use to test AI. One of our programs learned how to play 49 different Atari games from scratch just from seeing the pixels and score on the screen, and our AlphaGo program was the first to beat a professional player at the ancient game of Go, a feat experts described as a decade ahead of its time.