However, today’s chips take years to design, resulting in the need to speculate about how to optimize the next generation of chips for the machine learning (ML) models of 2-5 years from now. Dramatically shortening the chip design cycle would allow hardware to adapt to the rapidly advancing field of ML. What if ML itself could provide the means to shorten the chip design cycle, creating a more integrated relationship between hardware and ML, with each fueling advances in the other?
WHY IT MATTERS: always troubling when you first realize a computer can do your most difficult job better than you could.
I studied microelectronics in engineering school in the 1980s. I learned to create computer chips by placing and routing millions of components. It was the revolution of "silicon compilers", software that can place and route your design automatically instead of manually. But we had to write the program to control the compiler.
Bye-bye program and compiler it seems, as AI can now place and route billions of components better and faster than humans can. Oh, you are right, it is now possible for a computer to build another computer. And this is one more step towards the singularity...