Google claims that it has developed artificial intelligence software that can design computer chips faster than humans.
A chip that would take humans months to design could be dreamed up by its new AI in less than six hours, the tech giant said in a paper in the journal Nature on Wednesday.
Google said AI has already been used to develop the next iteration of Google’s Tensor Processing Unit chips, which are used to drive AI-related tasks.
“Our method has been used in production to design the next generation of Google TPUs,” wrote the paper’s authors, led by Azalea Mirhosini and Anna Goldie, Google’s co-heads of machine learning for the system.
To put it another way, Google is using AI to design chips that can be used to build even more sophisticated AI systems.
Specifically, Google’s new AI can create a “floorplan” of a chip. This essentially involves plotting where components such as the CPU, GPU and memory are placed on a silicon die in relation to each other – their position on these miniscule boards is important because it affects the power consumption and processing speed of the chip.
It takes humans months to optimally design these floorplans, but Google’s deep reinforcement learning system—an algorithm trained to take certain actions to maximize their chances of earning a reward—makes it relatively easy. Can do it with little effort.
Similar systems can even beat humans in complex games like Go and chess. In these scenarios, the algorithm is trained to move pieces that increase the chances of winning the game but in the chip scenario the AI is trained to find the best combination of components to make it as computationally efficient as possible. Can go A 10,000 chip floorplan was fed to the AI system to “learn” what worked and what didn’t.
While human chip designers typically lay out components in neat lines, Google’s AI uses a more scattered approach to design its chips. This isn’t the first time an AI system has gone haywire after learning how to perform a task behind human data. DeepMind’s famous “AlphaGo” AI took a highly unconventional move against Go world champion Lee Sedol in 2016 that astonished Go players around the world.
Google’s engineers noted in the paper that the breakthrough could have “major implications” for the semiconductor sector.
An editorial in Nature on Wednesday praised the success as a “significant achievement” that would be a “huge help to speed up the supply chain”.
However, the magazine said, “technical expertise must be widely shared to ensure that the ‘ecosystem’ of companies is truly global.” It emphasized that “the industry must ensure that time-saving technologies do not turn people away from essential core skills.”
Explanation: This story has been updated to reflect that Anna Goldie is a co-author of the paper, and that AI has been used to develop the next iteration of Google’s Tensor Processing Unit chips.
Read Original Article at www.cnbc.com