Li Guojie, an academician of the Chinese Academy of Engineering, shared his views on artificial intelligence computing chips at the Academician Forum of Science and Computing Microelectronics in China on Nov 15 during the 25th China Hi-Tech Fair in Shenzhen.

 

He said that computing power and chips have attracted special attention in the past two years, and new technologies represented by generative AI have pushed AI to a new height.

 

The improvement of computing power depends on the chip and system structure, Li added. Over the past half century, about half of the performance improvement of computer systems comes from microelectronics, or Moore's Law, and the other half comes from system structure and software. "In the next 20 years, the contribution of system structure innovation is likely to be greater than that of chips."

 

 

Academician Li Guojie delivers a speech at the forum. [Photo/Shenzhen Economic Daily]

 

"However, the problem is that Moore's Law has come to an end. The progress of semiconductor devices themselves and the development speed of supercomputers, which increases 1,000 times every 10 years, are far less than the demand for computing power, which is the big problem encountered by today's information technology," Li said.

 

"So new microelectronic devices to solve the computing power needs are imminent," he added.

 

The function of microelectronic devices is not limited to increasing computing power, and its complex systems may emerge with unimaginable functions like storage, memory, association and control, which are just as important as computing power.

 

Li also noted that analog computation might be another way to increase the computing power of chips.