How do AI work faster than us humans? How is it so accurate? Can it end the global chip crisis?
We are currently facing a huge shortage of semiconductors and electronic chips as the digital age progress faster. However, as we face this acute crisis, a small team of Google researchers is working on the design of the next generation of artificial intelligence (A. I. ) chips. Moreover, they have created an AI model that allows the chip design to be manufactured by using these artificial agents. The agents with artificial intelligence contain more experience than the average human designer.
The Google team used a data set of more than 10,000 chip layouts. They then fed the data to use it for machine learning and then trained them for proper AI with reinforcement learning. The new AI method for this research uses its experience to become much more accurate and precise oriented. This makes the machines adept at solving new instances of the problem.
The researchers used a method unique to Google. The research is used to design Google’s artificial intelligence (AI) accelerators. It has the potential to save thousands of hours of human work for each new generation. The current team of researchers believes in this cause and have elaborately explained the study in the scientific journal Nature.
In six hours, the AI model can generate a concrete design that optimises different parts of the chip.
The reinforcement learning in this method generates chip layouts in just a few hours. However, human experts can take months to make the same product. This was explained by Anna Goldie in a tweet. She is a research scientist at Google Brain and is also a part of this study.