IBM's Photonic Tensor Core can create Ultra-fast Computing in AI Systems

Quantum computers have the potential to change the way humankind design new materials, work out logistics, build artificial intelligence (AI) and break encryption. That is why firms like Google, Intel and IBM, along with plenty of start-ups have been racing to achieve this crucial milestone. And IBM researchers have recently unveiled a new approach that could bring big changes for deep-learning applications, processors that perform computations with light entirely, instead of electricity.

The researchers have created a photonic tensor core, based on the properties of light particles can process data at unprecedented speeds to deliver AI applications with ultra-low latency.

Let's explore how IBM is using light instead of electricity to develop ultra-fast computing in AI systems.

The device has only been tested on a small scale, but the report indicates that as the processor develops, it could achieve one thousand trillion multiply-accumulate (MAC) operations per second and per square-millimetre.

IBM has been working on novel approaches for processing units for several years now. Part of the company's research has focused on developing in-memory computing technologies, in which memory and processing co-exist in some form. It avoids transferring data between the processor and a separate RAM unit, saving energy and reducing latency.

Last year, IBM researchers developed an all-optical approach to in-memory processing. They incorporated in-memory computing on a photonic chip that used light to carry out computational tasks. The team demonstrated that a basic scalar multiplication could effectively be carried out leveraging the technology.

In a new blog post, Abu Sebastian, IBM researcher, shared a new milestone that has now been achieved using light-based in-memory processors. Taking the technology to the next stage, the team has built a photonic tensor core, a type of processing core that platforms sophisticated matrix math, and is well-suited for deep learning applications. The light-based tensor core was used to bring an operation called convolution that is useful to process visual data like images.

The most significant advantage that light-based circuits have over their electronic counterparts is never-before-seen speed. Using optical physics, the technology developed by IBM can run complex operations in parallel in a single core. Combined with in-memory computing, IBM's scientists achieved ultra-low latency that is yet to be matched by electrical circuits. For applications that require very low latency, hence, the speed of photonic processing could make a big difference.

With its ability to perform several operations simultaneously, the light-based processor developed by IBM also requires less compute density. It could be another key differentiator where loading car trunks with conventional GPUs to support even-more sophisticated AI systems will not cut it anymore.

With leading car companies now opening their own AI research centres, IBM sees autonomous vehicles as a key application for light-based processors. There is a real requirement for low latency inference in the domain of autonomous driving, and no technology that can meet it as of now. That is a unique opportunity.

Although IBM's researchers have successfully designed and tested a powerful core, they need to extend trials to ensure that the technology can be incorporated at a system level to ensure end-to-end performance. As per the scientists, the work is already underway, and as research continues, more applications are only likely to arise. Trading electricity for light, in the field of computing, certainly makes for a spot to watch.