Research by Anna Goldie and Azalia Mirhoseini has led to the development of AlphaChip, an AI method that accelerates and optimizes chip design. This groundbreaking approach has been used in the last three generations of Google’s Tensor Processing Unit (TPU), generating superhuman chip layouts in record time. AlphaChip uses reinforcement learning to automate the complex chip floorplanning process, similar to how AlphaGo master games like Go. The impact of AlphaChip extends beyond Google, influencing chip design across Alphabet and industry leaders like MediaTek. This innovative AI method has sparked a new wave of research on using reinforcement learning for chip design, promising faster, cheaper, and more energy-efficient chips in the future.
https://deepmind.google/discover/blog/how-alphachip-transformed-computer-chip-design/