Sign in

Optalysys

Since the early days of the silicon revolution, the course of processor development has followed a spectacular yet predictable arc. This prediction, the promise that the chips of tomorrow will be vastly more powerful than the ones of today, has for the last 50 years been encapsulated by Moore’s law, the observation that the density of transistors on a chip doubles roughly every two years. This persistent and reliable exponential growth in computational power has fuelled staggering global change and a correspondingly vast market for computing hardware. However, the promise of tomorrow is no longer the same…

Conventional computing is…


Over the past half year at Optalysys, we’ve been testing the prototype of our novel Fourier-optical processor technology in a broad range of applications. At the core of our concept is the fusion of silicon photonics with free-space optics, a technique that aims to couple the incredible computational capabilities of Fourier optics to the immense data throughput that can be achieved with modern integrated photonic components.

Thus far, every time we have put our technology to the test in a different application, it has yielded remarkably tractable results that have allowed us to successfully apply the system across radically different…


Introduction

Modern CFD techniques give us impressive capabilities when it comes to designing things like safer aircraft, more efficient wind turbines and better streamlined cars, but they also come at an incredibly high cost in money, time and energy. This is because we have to repeatedly solve vast numbers of equations representing small changes in flow properties over time. We asked if there was a better way of doing this.

There is.

In our last article, we gave a brief introduction to the field of computational fluid dynamics (CFD) and described one of the most popular methods for solving fluid flow…


The Fourier transform is a phenomenally useful mathematical tool applied throughout the scientific world. Wherever nature contains a repeating pattern, the Fourier transform can be used to break that pattern down, converting the information into a form which is much easier to work with.

However, some of these patterns are quite large, and only fully reveal themselves over many data points. The optical method of computing the Fourier transform is extremely fast, but such systems will always have some physical limits to the number of data points they can work with at once. …


By Florent Michel, Edward Cottle, and Joseph Wilson

Introduction

Transformer networks are an incredibly powerful deep learning architecture and they can be truly enormous. GPT-3, the natural language model that made headlines last year both for its exceptional performance and vast number of parameters, is a transformer network. However, training these networks takes up vast amounts of computing time, which in turn consumes a significant amount of power. As of 2019, GPU-powered training of a single large transformer network with neural architecture search could produce upwards of 625,155 lbs of CO2. …


If you’ve been reading our previous articles in this series about the workings of our optical computing system, we’ve been talking about how we can use our technology to set up an optical field that carries numerical information, and then process that optical field with a lens in order to calculate the Fourier transform of that data. …


In our last article, we explained how we use silicon photonics to control and modulate light with the aim of creating an optical field that contains data we want to process.

The Optalysys approach is very different to the other silicon photonic computing systems which are under development. Like our own designs, these systems use arrays of Mach Zehnder interferometers (MZIs) embedded in silicon to control light and perform multiplications. However, these other optical computing systems are only designed to perform multiplications for the Multiply and Accumulate (MAC) operations that are part of executing many deep learning models. They use…


In our first article, we talked about the basic ideas behind our optical Fourier Transform chip and why we made it a reality. In the next few articles, we’ll be going into greater depth on our system and how it works. We begin here by describing how we can take digital data and encode it into an optical field that can be processed at the speed of light.

Why use light?

Our technology is designed to perform a mathematical function called a 2-dimensional Fourier transform at very high speed. In our system, this calculation is performed through a combination of optical interference and…

Optalysys

We are developing Optical AI chips which will provide previously unseen levels of processing whilst consuming a fraction of the power of electronic processors.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store