Q/C Technologies Launches New Optical Processing Unit Initiative
Q/C Technologies announced the launch of a new initiative to design and prototype a proprietary optical processing unit aimed at tackling the growing performance and energy constraints in artificial intelligence inference infrastructure. Optical computing overcomes challenges posed by electronic GPUs. Artificial intelligence is dominated by one operation: the matrix multiplication. Photons can be configured to perform matrix multiplication naturally via interference achieving O(n) or even O(1) propagation latency, while most electronic devices used today, including GPUs, require O(n3) operations of work per matrix. Energy requirements of passive photonic components are minimal, addressing a key concern of electronic GPUs today. Photonic computing adds other advantages, including orders of magnitude faster clock speed and bandwidth. Q/C aims to overcome the three key limitations of optical computing: analog resolution, nonlinearity and storage via patentable proprietary technology and partnerships. "Q/C believes the future of artificial intelligence hardware is optical. With AI driving the next inflection point in computing architecture, we plan to develop our own optical chip that we believe can address key bottlenecks in bandwidth, energy efficiency, and scalability," said Executive Chairman Joshua Silverman. "We are encouraged by the caliber of AI, photonics and computing talent already engaged with Q/C including our newest director Chelsea Voss and strategic advisors Martin Shkreli and James Altucher. Together, they will spearhead a best-in-class solution for powering the future of AI."
