Researcher handling components for new AI memory device.
The AI industry’s voracious appetite for energy may soon be curbed, thanks to a new artificial memory device developed at the University of Cambridge. The device promises to drastically reduce AI energy consumption by up to 70%.
The technology hinges on a thin oxide film that operates with currents below 10 nanoamps and can hold hundreds of stable states. This innovation integrates data storage and processing, mimicking the human brain’s efficiency.
According to the research, the device exhibits smooth, controlled resistance changes, which are crucial for AI learning tasks. It also maintains stability over 50,000 switching cycles, retaining programmed states for approximately a day.
The core challenge lies in mass production. The high temperature required for film growth (near 1,292 degrees Fahrenheit) is incompatible with standard chip-making processes. However, if this hurdle can be overcome, these artificial memories could enable dense AI arrays that learn and operate efficiently in place, reducing the substantial energy cost associated with data transfer in current AI systems.
The study was published in Science Advances.
About the Author
