Developed by Google AI Capable of creating computer chips in less than six hours, the research, published in Nature, suggests that humans can take months to design. Specialized slides For tensor processing units, a type of chip used in artificial intelligence, but the reinforcement learning (RL) algorithm is better and faster than humans at creating more complex AI.
According to the British newspaper, “Daily Mail”, the researchers wrote in the study: “The algorithm becomes better and faster in optimizing floor planning as it creates a greater number of network lists.”
“We show that our method can generate chip patterns that are comparable to or superior to human experts in as little as six hours, while it takes humans months to produce acceptable patterns for modern accelerators,” the researchers added.
It’s already being used commercially in the latest version of CPU chips, Google researchers said. The software researchers gave 10,000 chip diagrams to analyze, then figured out how to come up with blueprints that don’t use more space, wires and electrical power than those designed by humans.
The chip outline is where parts such as CPUs GPUs and memory on silicon, and AI could have broader implications for the semiconductor industry.