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

I think recently RL or DL is going crazy with design. This is very nice to see.

https://www.nature.com/articles/d41586-021-01515-9