Using Light to Process 3D Data – A New Breakthrough in AI Hardware

Friday - October 20 2023, 21:25 UTC - 1 year ago

Scientists from Oxford University, Germany, and the UK developed a new AI hardware using light to process three-dimensional (3D) data. This chip is faster and more efficient than traditional electronic chips and is expected to have even more potential with more features. It was tested on a real-world problem to detect risk from heart signals with 93.5% accuracy.

A team of scientists from Oxford University and their partners from Germany and the UK have developed a new kind of AI hardware that uses light to process three-dimensional (3D) data. Based on integrated photonic-electronic chips, the hardware can perform complex calculations in parallel using different wavelengths and radio frequencies of light. The team claims their hardware can boost the data processing speed and efficiency for AI tasks by several orders of magnitude.

AI computing and processing power .

The research published today in the journal Nature Photonics addresses the challenge of meeting modern AI applications' increasing demand for computing power. The conventional computer chips, which rely on electronics, need help to keep up with the pace of AI innovation, which requires doubling the processing power every 3.5 months. The team says that using light instead of electronics offers a new way of computing that can overcome this bottleneck.The team had previously demonstrated a prototype of a photonic chip that could perform matrix-vector multiplication, an essential operation for AI and machine learning, faster than any electronic chip. This chip used different light colors to represent different data sets and carry out multiple calculations in parallel. This breakthrough led to Salience Labs, a spin-out company from Oxford University focusing on photonic AI.

Photonic chip .

In their latest work, the team has added another dimension of parallelism to their photonic chip by using different radio frequencies of light to encode the data. With much higher speed and accuracy, the chip can process 3D data, such as images or videos. The team tested their hardware on a real-world problem of predicting the risk of sudden death from heart signals in patients. They could analyze 100 heart signals simultaneously and identify the trouble with 93.5% accuracy.

The team estimates that their hardware can outperform the best electronic processors by a factor of 100 in terms of energy efficiency and compute density, even with a modest scaling of 6 inputs × 6 outputs. They also expect to achieve further improvement in computing parallelism by exploiting more light features, such as polarization and mode multiplexing.

Dr. Bowei Dong, the first author of the paper and a researcher at the Department of Materials, Oxford University, said: "We used to think that using light instead of electronics could only increase parallelism by using different colors – but then we discovered that using radio frequencies to represent data opens up a whole new dimension, enabling superfast parallel processing for emerging AI hardware." .

Professor Harish Bhaskaran, the leader of the research and the co-founder of Salience Labs, said: "This is an exciting time to be doing research in AI hardware at the fundamental scale, and this work is one example of how we can push the boundaries of what we thought was possible." .