Researchers have come up with a photonic processor they claim could take some of the ‘sting’ out of the conventional technology that currently drives AI.
The deep neural network models that power today’s most demanding machine-learning applications have grown so large and complex that they are pushing the limits of traditional electronic computing hardware.
Meanwhile, the massive rise in data centres, driven by the growth of AI/machine learning systems and cloud computing, are causing issues of their own: pressures on water resources for cooling, high energy demand and the resulting carbon emissions, for example.
But the new device developed by researchers at the Massachusetts Institute of Technology (MIT) uses light to perform the key operations of a deep neural network on a chip. It opens the possibility of a new generation of powerful computer processors that are faster – vastly more energy efficient – and can learn in real time.
Up to now, there are some types of neural network computations that a photonic device can’t perform, requiring the use of off-chip electronics or other techniques that hamper speed and efficiency.
However, building on a decade of research, scientists from MIT and elsewhere have developed the new photonic chip that overcomes these roadblocks. They demonstrated a fully integrated photonic processor that can perform all the key computations of a deep neural network optically on the chip.
The optical device was able to complete the key computations for a machine-learning classification task in less than half a nanosecond while achieving more than 92% accuracy – performance that is on par with traditional hardware.
The chip, composed of interconnected modules that form an optical neural network, is fabricated using commercial foundry processes, which could enable the scaling of the technology and its integration into electronics.
In the long run, the photonic processor could lead to faster and more energy-efficient deep learning for computationally demanding applications like lidar, scientific research in astronomy, and particle physics, or high-speed telecommunications.
Saumil Bandyopadhyay is a visiting scientist in the Quantum Photonics & AI Group within the Research Laboratory of Electronics (RLE), and a postdoc at NTT Research, Inc.
Also serving as the lead author of a paper on the new chip, Bandyopadhyay said: “There are a lot of cases where how well the model performs isn’t the only thing that matters, but also how fast you can get an answer. Now that we have an end-to-end system that can run a neural network in optics, at a nanosecond time scale, we can start thinking at a higher level about applications and algorithms.”
The research appears in the journal, Nature Photonics.
Image credit: Sampson Wilcox, Research Laboratory of Electronics. Courtesy of MIT
Read next: Project partners celebrate completion of Leeds flood defences
Are you a building professional? Sign up for a FREE MEMBERSHIP to upload news stories, post job vacancies, and connect with colleagues on our secure social feed.
