Google has developed an artificial intelligence program that can design computer chips faster than humans, using machine learning to help design the next generation of machine learning chips.
The new system from “Google” can develop a diagram of the chip includes layout, where components such as: CPUs, GPUs, and memory, are placed through a silicon mold in relation to each other.
In a research study, engineers from Google said that designs for the algorithm called “Reinforcement Learning” are comparable to, or superior to, those designed by humans, and can be created much faster.
They added that the chip, which could take humans months to design, could be devised by the new AI in less than 6 hours.
The company indicated that artificial intelligence was used to develop the latest version of “Google’s Tensor” processing unit chips.
It takes humans months to design these schemas optimally, but Google’s deep reinforcement learning system can do so with relatively little effort.
The system was fed with about 10,000 chip blueprints in order to examine what is working and what isn’t, knowing that Google’s AI takes a more dispersed approach to designing its chips.
Google engineers noted, in the research paper, that this breakthrough could have major implications for the semiconductor sector.