Tag Archives: #google#AI#processors#manafacturing

Google’s next generation AI processor, which will succeed the TPU version 4, was created in part using an AI that researchers detailed to IEEE Spectrum last year, according to a paper published this week in the journal Nature. Since the last time Spectrum spoke with them, they’ve made some improvements. In less than six hours, the AI can build chip floorplans that are comparable to or better than human-produced designs in terms of power consumption, performance, and area. It takes months of iteration for an expert human to do this work.There has been a lot of hard effort and money put into building chips that are specifically intended to conduct AI algorithms more quickly and efficiently. The problem is that designing a semiconductor takes years, whereas the world of machine learning algorithms moves much faster. You want a chip that is tuned for today’s AI, not two to five years ago’s AI. Google’s solution: having the AI chip designed by an AI.

Google is leveraging AI to create more efficient AI processors. Google has a massive computational infrastructure.
Google engineers gave an artificial intelligence the duty of building faster and more efficient processors, then used the chip designs to create the next generation of specialised computers that run the internet.