Google used reinforcement learning to design next-gen AI accelerator chips

In a paper published in the journal Nature, Google detailed its method for creating an AI system capable of designing computer chips.