As humans increasingly use artificial intelligence technologies, the energy required to power AI data centers is a concern. Now, University of Missouri researchers are studying how to make computing more energy efficient.
Most electronic devices expend energy by first processing a piece of information and then transferring it to storage in a separate unit.
MU physics professor Suchi Guha researches electronic devices and is looking into how to increase computer efficiency by making them work more like a brain.
“In a brain, the processing and the memory unit are all together, so you don't have this data exchange between the CPU and the memory, making it much, much more efficient,” Guha said.
Guha and her research team are developing what’s called “neuromorphic hardware” — a flexible computer chip prototype that can complete basic computing tasks such as pattern recognition and sensing.
Energy demand is expected to skyrocket due to the build out of AI data centers. Missouri utility Ameren plans to increase the power supply by 50% in the next four years to meet growing demand.
But Guha said all the electronics in those data centers use traditional computing.
“One of the brain’s greatest advantages is its efficiency,” she said in a university report. “It performs incredibly complex tasks using about 20 watts of power — roughly the same as an old light bulb. By comparison, today’s computer architecture is extremely energy-intensive.”
Guha said the research to make computers more efficient, to process and store information simultaneously, is in a rudimentary stage. But, if successful, her team estimates the new computer chips would require just 4 percent of the power used in current electronics.
