One of the biggest forces shaping the future is artificial intelligence (AI). DeepMind, Google's AI research lab based here in London, is at the forefront of this research. We went and spoke to Alex Graves, research scientist at DeepMind, about their Atari project, where they taught an artificially intelligent 'agent' to play classic 1980s Atari videogames.
After just a few hours of practice, the AI agent can play many of these games better than a human. Using machine learning, a process of trial and error that approximates how humans learn, it was able to master games including Space Invaders, Breakout, Robotank and Pong. While this demonstration may seem trivial, it is the first example of flexible intelligence – a system that can learn to master a range of diverse tasks. And as Alex explains, it points toward research to address grand human challenges such as healthcare and even climate change.
