Nvidia: It only takes a few days for two AI GPUs to design a chip better than 10 people working hard in a year

The development of artificial intelligence (AI) technology is bringing people countless practical benefits, but at the same time, it also brings concerns. On the one hand, people feel excited about the achievements and contributions of AI in all aspects of life. But on the other hand, many people are "fearful" about the prospect that AI can "rob" people of jobs in many different fields.

Nvidia is a big "player" in the semiconductor manufacturing industry, and at the same time is also one of the pioneers in the field of research and application of AI in practice. Therefore, it is not surprising that the company is making AI systems that design chips that are significantly faster and more efficient than humans.

According to a recent statement from Nvidia's Senior Vice President of R&D, engineer Bill Dally, the company began designing the chip using an AI-based automatic standard cell layout generation system called NVCell. This system not only helps to check and detect errors in the chip layout, but it can also take care of the processor design process at a much faster speed than a human. Nvidia says it only takes a few days for the company's two AI GPUs to do 92 percent of the work a 10-person team should complete in a one-year period.

Also, one of the other big advantages of AI is its willingness to take risks with designs that humans would normally not dare to consider. For example, designers often intentionally leave a certain amount of error when placing circuits on chips to "prepare" errors in the manufacturing process. Normally, engineers tend to want to keep the error range as wide as possible, but AI is the opposite.

 
Nvidia isn't the only major manufacturer impressed by AI's chip design capabilities. In June of last year, Google also announced a similar project with an impressive performance of AI in the aspect of creating chip layout. Google Brain - Google's AI and deep learning team - has announced the successful development of a new reinforcement-learning system that can complete a floor-planning design on its own. ) of microprocessors much faster and better than humans.

With the help of a complex neural network architecture based on edge graphs, Google Brain's AI model can design its own diagrams of floorplans in a fraction of the time it takes a human.

Hopefully AI can make remarkable breakthroughs in chip fabrication in the near future.

NVIDIA unveils technology that can turn 2D photos into 3D scenes in the blink of an eye
AI in war: "Suicide" UAVs begin to appear on the battlefield
China tests its first artificial intelligence prosecutor
Based on GPT-3, they have developed a tool that can turn your commands into lines of code