Artificial intelligence systems, including large language models and image generators, consume large amounts of electricity. Engineers at the University of Wisconsin–Madison have designed a new type of quantum nanostructure that could address this problem by making optical neural networks more feasible.
Optical neural networks process information using light instead of electronic signals, which could be far more energy-efficient than conventional hardware. The newly designed nanostructures are a step toward building such networks in practice, according to the researchers.
The work was reported by Phys.org. The team says the technology has the potential to make AI less of an energy burden, though the report does not specify a timeline for commercial deployment.