Google Chief Scientist: We have started to produce custom chips that accelerate AI

At present, AI has become a point of promotion for many smart phones, such as AI face recognition, AI chips, AI scene recognition, etc. When it comes to products equipped with artificial intelligence chips, perhaps everyone will think of the Unicorn 970's NPU processing unit. A11 bionic processor neural network or Qualcomm 660 AIE...

In fact, Google is actually a pioneer in terms of artificial intelligence and artificial intelligence chips. I don't know if everyone still remembers TPU, a Google-built processor that is tailor-made for machine learning.

▲ Google's second-generation TPU

So, what does Google think of AI chips?

At the AI ​​technology sharing conference held in the Chinese office, Google’s chief scientist Greg Corrado gave his opinion on the chip issue. He believes that Google has now begun to produce custom chips that can accelerate AI, but such chips are not dedicated. AI chip.

At the same time, he said that so far he has not seen any successes that are completely different from traditional computing chips. He believes that existing chips should be optimized specifically for AI, making the chips faster and lower power consumption when completing AI tasks. The benefits are even higher.

Greg Corrado uses AlphaGo as an example. He thinks that AlphaGo's computing chip is Google's optimized chip for AI.

2016 GoodChinaBrand | ICP: 12011751 | China Exports