University of Pennsylvania

Bīnxīfǎníyà Dàxué

宾夕法尼亚大学

researchers

yánjiū rényuán

研究人员

to make use of / to harness

lìyòng

利用

light-matter

guāng wùzhì

光物质

hybrid

hùnhé

混合

particle

lìzǐ

粒子

significantly / by a large margin

dàfú

大幅

to improve / to boost

tíshēng

提升

artificial intelligence

rén gōng zhìnéng

人工智能

computing / calculation

jìsuàn

计算

efficiency

xiàolǜ

效率

。

New Words

News in English

Penn Researchers Harness Light-Matter Hybrid Particles to Boost AI Computing Efficiency

Physicists at the University of Pennsylvania have created a hybrid light-matter particle that enables AI optical signal switching at extremely low energy, potentially replacing conventional electronic computing steps.

University of Pennsylvania physicists have published their latest research findings in Physical Review Letters.

The research team directed light into nanoscale cavities, causing it to interact with atomic-scale thin materials, thereby producing exciton-polaritons.

These hybrid particles combine the speed of light with the interactivity of matter, enabling optical signal switching.

During testing, this switching operation consumed only about 4 femtojoules of energy, far lower than existing electronic devices.

Currently, many photonic AI chips still need to convert optical signals back to electronic signals when performing nonlinear activation steps, significantly reducing the efficiency advantages of photonic computing.

The novel exciton-polaritons technology promises to solve this bottleneck, enabling the entire computing process to operate at light speed.

The project is led by University of Pennsylvania physicist Bo Zhen, whose team focuses on photonic and quantum materials research.

More in Science & Technology