Creative rendering of a photonic chip with both light and RF frequency encoding information. Credit: B.Dong/ University of Oxford
Researchers have actually developed an incorporated photonic-electronic hardware capable of processing 3D information. This innovation considerably enhances data processing parallelism for AI tasks.
An advancement in photonic-electronic hardware might significantly enhance processing power for AI and artificial intelligence applications.
The technique utilizes multiple radio frequencies to encode data, allowing numerous computations to be brought out in parallel.
The method shows guarantee for outperforming advanced electronic processors, with additional enhancements possible.
New Advancements in Photonic-Electronic Hardware for AI
In a paper published on October 19 in the journal Nature Photonics, scientists from the University of Oxford, in addition to collaborators from the Universities of Muenster, Heidelberg, and Exeter, report on their development of integrated photonic-electronic hardware efficient in processing three-dimensional (3D) data, substantially increasing data processing parallelism for AI jobs.
Difficulties With Current Computing Power and the Role of Photonics
Conventional computer system chip processing effectiveness doubles every 18 months, however the processing power required by contemporary AI jobs is currently doubling around every 3.5 months. This indicates that new computing paradigms are urgently needed to deal with the increasing demand.
Artistic rendering of a photonic chip with both light and RF frequency encoding data. Credit: B.Dong/ University of Oxford.
One method is to utilize light instead of electronic devices– this permits multiple estimations to be performed in parallel using different wavelengths to represent various sets of data. Undoubtedly, in groundbreaking work published in the journal Nature in 2021, much of the exact same authors showed a type of integrated photonic processing chip that might bring out matrix vector reproduction (an important task for AI and artificial intelligence applications) at speeds far outmatching the fastest electronic methods. This work led to the birth of the photonic AI company, Salience Labs, a spin-out from the University of Oxford.
Innovations in Parallel Processing and Real-world Application
Now the group has actually gone further by including an extra parallel measurement to the processing ability of their photonic matrix-vector multiplier chips. This “higher-dimensional” processing is enabled by exploiting several different radio frequencies to encode the information, propelling parallelism to a level far beyond that previously achieved.
As a test case, the group applied their novel hardware to the job of assessing the danger of unexpected death from electrocardiograms of heart problem clients. They were able to effectively analyze 100 electrocardiogram signals simultaneously, determining the risk of unexpected death with a 93.5% accuracy.
Future Prospects and Expert Opinions
The scientists further approximated that even with a moderate scaling of 6 inputs × 6 outputs, this method can surpass cutting edge electronic processors, potentially offering a 100-times enhancement in energy effectiveness and compute density. The group prepares for further enhancement in computing parallelism in the future, by exploiting more degrees of freedom of light, such as polarization and mode multiplexing.
Author Dr. Bowei Dong at the Department of Materials, University of Oxford said: “We formerly presumed that utilizing light rather of electronic devices could increase parallelism only by the use of various wavelengths– but then we realized that utilizing radio frequencies to represent data opens up yet another measurement, enabling superfast parallel processing for emerging AI hardware.”
Professor Harish Bhaskaran, Department of Materials, University of Oxford and CO-founder of Salience Labs, who led the work stated: “This is an amazing time to be researching in AI hardware at the basic scale, and this work is one example of how what we assumed was a limit can be further exceeded.”
Recommendation: “Higher-dimensional processing utilizing a photonic tensor core with continuous-time data” by Bowei Dong, Samarth Aggarwal, Wen Zhou, Utku Emre Ali, Nikolaos Farmakidis, June Sang Lee, Yuhan He, Xuan Li, Dim-Lee Kwong, C. D. Wright, Wolfram H. P. Pernice and H. Bhaskaran, 19 October 2023, Nature Photonics.DOI: 10.1038/ s41566-023-01313-x.