AI designs chips to enable more efficient AI, which will then innovate better AI chips, ad infinitum
As is so often the case with reinforcement learning, the AI goes rogue, here arraying chips in a scattered, non-linear fashion
Click here for related article