Scientists at the university have recently achieved a remarkable breakthrough in the realm of next-generation optical computing chips. For the first time, they have successfully developed a fully optical computing chip capable of supporting large-scale semantic media generation models. The pertinent research results were published in the esteemed journal Science on December 19.

The research team, headed by Chen Yitong, introduced and brought into reality LightGen, a fully optical large-scale semantic generation chip. Tests reveal that even when paired with lower-performance input devices, this chip can still deliver a computing power and energy efficiency improvement of two orders of magnitude compared to leading digital chips. This significant performance leap is credited to the overcoming of three major bottlenecks, paving the way for the full-optical, end-to-end accomplishment of large-scale generation tasks.