Photonic AI: exciting new developments
(c) Bowei Dong/University of Oxford
"This is an area which is moving rapidly with a lot of interesting science and engineering to explore".
In a new paper* published today* (31 July 2024) in Nature, a multi-institutional team of researchers (which includes The Bhaskaran Lab) demonstrate how less complex (and cheaper) light sources can be utilised for emerging light-driven applications - for example, optical AI technologies. In addition to widening the scope of sources which boost performance, less energy will also be required than conventional high-specification lasers.
The team used a partially coherent light source by harnessing a narrow portion of the spectrum of incoherent light produced by an electrically-pumped erbium-doped fibre amplifier (EDFA). This partially coherent light was evenly split and distributed into different input channels for a parallel AI computational array. Using such a light source, the parallelism of AI computation is surprisingly enhanced by N times in a photonic accelerator with N input channels.
Ultimately, removing the need to add additional light sources could prove transformational in boosting computational power. First author Dr Bowei Dong explains that "the benefit of using 'poorer' light sources has a scaling effect. You can run your AI models 100 times faster compared to a laser system, if the photonic accelerator scales to 100 input channels".
Read more about this exciting new development on the University of Oxford's webpage: 'New 'game-changing' discovery for light-driven artificial intelligence'.
*'Partial coherence enhances parallelized photonic computing'.
