Artificial intelligence can design computer microchips that work at least as well as those designed by humans.
AI is better than other automated methods “because it learns from experience”
The advantage of artificial intelligence is that it can design such designs thousands of times faster. This new Google research is already helping to design microchips for the company’s next generation of AI computing systems. The process of designing the physical layout of chip parts, known as basic planning, is key to the ultimate performance of a device. This complex task often requires months of intensive professional effort. In the new study, researchers programmed AI to treat basic planning as a game, said study co-author Azalia Mirhoseini, a scientist from the Google Research Brain Team.
In this approach, the computer core of the chip, memory controllers and other elements are parts, the screen on which they are located is the board, and the winning result is optimal performance in accordance with a number of values, such as chip size and power consumption. As AI analyzed the greater number and variety of microchips, it became faster and better at designing chips. This strategy has proven to be better than previous automated techniques “because it learns from experience,” said study co-author Anna Goldie, a scientist also from Google’s Brain Team. Previous approaches have not learned from examples.
In less than six hours, the new method can generate workable basic chip plans, which are comparable or superior to those made by experts in all key details, including performance, power consumption, and the area the chip occupies. This could potentially save thousands of hours of human effort, for each generation of microchips. Some of these designs were not intuitively better at first, but they still brought really good results, Mirhoseini said. Artificial intelligence design has inspired experts for new ideas, such as placing donut-shaped components to reduce the distance between them.
Google now produces these chips designed by artificial intelligence. Researchers have suggested that more powerful AI-designed hardware can in turn encourage greater advances in artificial intelligence itself. The new strategy could also help address other issues that include finding the best use of limited sets of resources, such as city planning or testing and distributing vaccines, the scientists added. They published their findings in detail in the June 10 issue of the journal Nature.
Source: PC Press by pcpress.rs.
*The article has been translated based on the content of PC Press by pcpress.rs. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!
*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.
*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!