Intel Works for AI Friendly Optical Chips

Santa Clara based chipmaker Intel investigating AI workloads on silicon photonics. Their scientists at the company recently detailed in a paper about novel techniques that could bring optical neural networks a step closer to reality.

Photonic integrated circuits, or optical chips, can be exactly tailor-made for AI workloads because they offer a number of advantages over their electronic counterparts. These are about reduced power consumption and faster processing.  Some scientists, and that includes MIT Ph.D. candidate Yichen Shen , have founded companies to manufacture them commercially.

The co-authors explained in a blog post, that some earlier work has shown that a type of photonic circuit known as a Mach-Zender interferometer (MZI). This can be configured to perform a two-by-two matrix multiplication between quantities related to the phases of two light beams. When these small matrix multiplications are arranged in a triangular mesh to create larger matrices, they produce a circuit that implements a matrix-vector multiplication, a core computation in deep learning.

The scientists say their research could lay the groundwork for AI software training techniques that might eliminate the need to fine-tune optical chips post-manufacturing. This could save big on time and labor.

“As in any manufacturing process, there are imperfections, which means that there will be small variations within and across chips, and these will affect the accuracy of computations,” wrote Intel AI products group senior director Casimir Wierzynski. “If ONNs is to become a viable piece of the AI hardware ecosystem, they will need to scale up to larger circuits and industrial manufacturing techniques … Our results suggest that choosing the right architecture in advance can greatly increase the probability that the resulting circuits will achieve their desired performance even in the face of manufacturing variations.”