High-performance photonic AI chip design

A team of scientists at Tsinghua University in China has designed an artificial intelligence (AI) chip based on photonics instead of traditional semiconductor electronic components, paving the way for large-scale photonic computing and efficient AI applications in the real world (photo).

Accordingly, a large-scale photonic chiplet (a chip composed of many different blocks) along with a distributed optical computing architecture called Taichi has achieved high accuracy in complex classification tasks, such as the ImageNet 100-category dataset and the Omniglot 1,623-category dataset.

According to the research, this chip is also capable of performing high-precision tasks such as composing music and creating paintings. Furthermore, Taichi has achieved an energy efficiency of 160 tera operations per second per watt, a significant improvement in performance compared to current photonic integrated circuits and two steps above traditional AI chips in terms of energy efficiency.

According to expert Fang Lu at Tsinghua University, "We predict that Taichi will accelerate the development of more powerful optical solutions, such as crucial support for platform modeling and the new era of generative AI."

LAM DIEN

Source: https://www.sggp.org.vn/thiet-ke-chip-ai-quang-tu-hieu-suat-cao-post735709.html