Google has developed a unique software system based on artificial intelligence that can develop computer chipsets faster than humans. Basically, you can design a “floor view” of the chipset in just six hours, while for a human engineer, it takes months. Use new software to develop the next-generation Tensor data processing unit.
The Mountain View giant recently published a research paper on new chip design software in the journal Nature. The article stated that the company is using this artificial intelligence software to develop chipsets that can be used to create more advanced artificial intelligence systems.
How does it work?
Once you start using the new software, you can develop a “blueprint” for the chipset within six hours. The chipset plan is now the basic silicon-on-chip design that determines the best location for the CPU. , GPU and memory blocks on the same chipset.
For chip engineers, it usually takes several months, and Google’s new artificial intelligence can complete the design in six hours. The developer provided more than 10,000 floor plans to help the AI system investigate issues and find the best floor plan. A chipset that provides higher performance while consuming less power.
Use your knowledge to design a floor plan with a microchip. Although human engineers tend to create designs by arranging components in neat lines, Google’s artificial intelligence software takes a more decentralized approach. CPU, GPU, and memory blocks are in random locations until the design is complete.
Impact of New AI Software
After developing this AI software, the authors of this article, Azalea Mirhoseini and Anna Goldie, are also co-directors of Google’s machine learning system. They stated that their method “is used to produce the next generation of Google TPU” and will be used More complex chipsets will be developed in the future.
They also pointed out that these advances may have a significant impact on the semiconductor industry in the next few days, so the company can use its new artificial intelligence system to develop a series of chips that may be used in conjunction with the first Whitechapel chipset. From Google. -The mobile chipset for the next generation Pixel 6.