From Uber to Gmail’s spam filters, and smart-home devices like Siri, all our daily activities and apps rely on artificial intelligence. That said, executing these algorithms often require a huge quantity of computing power which produces large electricity usage and increased carbon footprints. Systems that minimise computer connections will enhance energy efficiency. This is now possible with the new computing device that scientists at Osaka University built. The researchers have built a field-programmable gate arrays (FPGA) computing device that can be adjusted by the user for maximum performance in AI applications.

Compared to the hardware currently used, the new system increases circuit density by a factor of 12. This can save energy consumption by up to 80 per cent. This evolution can lead to flexible artificial intelligence solutions that offer increased performance while using far less power.

The research team used non-volatile “switches” that remained connected until the user decided to reconfigure them. With the new nano-manufacturing process, they can package twelve times as many items in settings such as a grating as a cross belt.  

Typical hardware, which includes the physical logic gates and transistors of a computer’s processor, will be fixed by the manufacturer. That said, field-programmable gate arrays are specialised logical elements that can be rewired “in the field” by the user for custom logic applications. By minimising the distance in which the electronic signals are to be routed, the devices will use 80 per cent less power.

“Our system based on field-programmable gate arrays has a very fast design cycle. It can be reprogrammed daily if desired to get the most computing power for each new AI application,” the lead author in the research team Masanori Hashimoto says. “The use of via-switches also eliminates the need for the programming silicon area that was necessary for previous FPGA devices.”

Via-switch FPGA is suitable as a high-performance implementation platform of the latest AI algorithms.