Optical computing for more sustainable artificial intelligence
Computing & Tech
An Oxford university spin-out is developing faster and more efficient optical neural networks
Spotted: While most people are focused on the benefits of AI, very little has been said about its energy consumption. As AI models become more complex, the need for servers to process the models, and energy to run the servers, grows exponentially. A recent paper from the University of Massachusetts Amherst found that training a single AI model can emit as much carbon as five cars do in their whole lifetimes. In fact, today’s computers aren’t quick enough or energy-efficient enough for the demands of big data, and the future of AI.
This is where startup Lumai comes in. The company, which was spun out of Oxford University, is developing the next generation of AI – a generation that will be faster and more sustainable. These computers will use a new physical method for ‘training’ and ‘learning’ in AI. Rather than traditional electronic neural networks, Lumai’s system will use optical neural networks (ONNs). These use photons instead of electrons to transfer information and perform calculations, making them faster and more energy-efficient.
Lumai has already demonstrated the world’s largest optical matrix-vector multipliers, which will serve as the backbone of its ONNs. It is now developing schemes for training ONNs in an all-optical setting, which would be a world-first. To help make its tech a commercial reality, the company has recently received a £1.1 million (about €1.25 million) Innovate UK Smart Grant.
As computing demands more and more energy, innovators are increasingly focusing on ways to improve the sustainability of the industry. Some ideas that Springwise has recently spotted include using microfluidics to cool computers, and a fully recyclable computer chip substrate.
Written By: Lisa Magloff
27th February 2023