Researchers in the U.S. and China have collaborated to create a tiny microchip that carries out its calculations in memory instead of requiring separate chips for processing and data storage.
It’s twice as energy-efficient as today’s chips that combine memory and processing and delivers results that are just as accurate, the inventors say.
The chip is designed to carry out several functions in artificial intelligence (AI), including image and voice recognition, not only using a fraction of the energy that high-end computing usually demands but also carrying out its tasks without connecting to a central server the way AI often needs to.
Normally, AI requires large data servers operating in the cloud to maximize computing power and provide power to energy-hungry AI apps, which precludes operating on a single device not dependent on the cloud.
The developers envision their new NeuRRAM chip birthing a world of smarter wearables, smarter sensors all over factory floors, and more capable research robots landing on faraway planets, among a host of other applications.
TRENDPOST: Artificial intelligence scaled to the device level but retaining AI’s essential power and capacity is a formidable challenge in computing. The new chip and others like it will overcome that challenge and lead to a new generation of computer intelligence and power available to the rest of us as smartphones are now.