source: sciencedaily ai: forget electrons, this breakthrough uses light-matter particles to power ai
level: research
researchers at the university of pennsylvania have developed a hybrid light-matter particle called an exciton-polariton. it forms when photons couple strongly with electrons in an atomically thin semiconductor. this quasiparticle combines light's speed with matter's ability to interact, making it useful for signal switching in computing. traditional electronic chips rely on electrons, which generate heat and face resistance as they move. photons avoid these issues because they are charge-neutral and massless, but they normally struggle with the switching logic that computers need.
the team demonstrated all-light switching using only about 4 quadrillionths of a joule of energy. that is far less than what a tiny led needs to flash. many experimental photonic ai chips already use light for fast calculations. however, they often convert light back to electronic signals for nonlinear activation steps like decision-making. this conversion slows processing and increases energy use. the new approach could keep signals in the optical domain, reducing those losses.
if scaled up, the technology could lead to photonic chips that process information directly from cameras without repeated light-to-electricity conversions. it might also lower the massive energy demands of large ai systems and support basic quantum computing functions on future chips. the work was published in physical review letters and supported by the us office of naval research and the sloan foundation.
why it matters: this could reduce energy use and latency in ai hardware by enabling all-optical switching, a key step toward practical photonic computing.
source: sciencedaily ai: forget electrons, this breakthrough uses light-matter particles to power ai