A new kind of nanoelectronic device could dramatically cut the energy used by AI hardware through mimicking the human brain, according to a new study.
In the study led by the University of Cambridge, researchers developed a form of hafnium oxide that acts as a highly stable, low‑energy ‘memristor’. This electronic component acts as a resistor with memory, mimicking the way neurons in the brain form and adjust connections.
As AI adoption increases, this brain-inspired, or neuromorphic, computing could offer an energy-efficient way for AI systems to process information. Current AI systems rely on conventional computer chips that shuttle data back and forth between memory and processing units, utilising large amounts of electricity.
According to the researchers, their device could reduce energy use by as much as 70% by storing and processing information in the same place. It would also be more adaptable, in the same way our own brains are able to learn and adapt.
“Energy consumption is one...