How is Google DeepMind applying artificial intelligence to accelerate breakthroughs in clean baseload energy?

We are actively applying reinforcement learning algorithms to achieve magnetic confinement control of superheated plasma inside nuclear fusion tokamaks. Plasma behavior is notoriously turbulent and unstable, making real-time human control impossible. Our deep reinforcement learning agents adjust magnetic coil voltages thousands of times per second, moving us closer to commercially viable, limitless fusion power.