Researchers at the University of Tokyo have developed a novel computing architecture called diffraction casting, which aims to address the limitations of electronic computing by leveraging the power of optical computing. This innovative approach could pave the way for a new generation of high-speed, energy-efficient computers capable of tackling complex applications like artificial intelligence.
Harnessing the Power of Light
Most of today’s computer systems rely on electronics, with its characteristic limitations. As devices get more powerful, they create more heat, (shown to the lower right) and we’ve burned up most of the underlying technology in use now. Over the last several decades, researchers have been investigating using light waves to make computers work faster and more efficiently.
Optical computing as an idea has been floating around for the better part of a few decades now, but up to this point it still simply doesn’t provide a realistic alternative to standard electronic methods. “In the 1980s, researchers in Japan investigated an optical computing method known as shadow casting [which can carry out a few basic logical operations],” said Associate Professor Ryoichi Horisaki from the Information Photonics Lab at Tokyo University. Instead, these were implemented in voluminous geometric optical shapes, kind of like the vacuum tubes of early digital computers. They were working at most platitudes, but not flexible enough to be truly useful.
Diffraction Casting is born
In order to achieve optical computing with more performance than previous methods, the University of Tokyo team has developed a new design architecture which they named diffraction casting. ‘In diffraction casting, we improve from shadow casting,’ says Horisaki. Shadow-casting operates the way I quoted Crick above — light rays interacting with different geometries,while diffraction casting is based on properties of the light wave itself, resulting in optics that are spatially dense and can more easily embody requisite extensibility you’d expect and require for a universal computer.
The team has run numerical simulations that have produced encouraging results starting with small 16-by-16 pixel black-and-white images, which are smaller than the icons on a smartphone screen. “Numerical simulations confirmed that this induced very good results,” says Horisaki. ” We take an image as data source here—this would of course imply this system can be used for image processing, but other kinds of (machine learning) data could also be visualarily processed—but we use a set of other images which represent stages in doing logic operations on that source.
The Path to Commercialization
Diffraction casting would not replace all of our existing electronic systems, but it could one day serve as a brick in an ultra-fast computer system. “Diffraction casting is just a piece of the puzzle in a computer concept based on such an idea and it should be considered as another component or specialized system rather than replacing already existing systems; likewise GPUs are specialized for graphical processing, gaming and machine learning workloads,” says lead author Ryosuke Mashiko.
The idea is based on actual science, but the real-world application has a long way to go before it can be bought at a store. “Commercially, I expect it to come around in the next 10 years that were guesstimates before you can actually synthesize it because there is so much more engineering and machinery work although the idea itself does work [she laughs] but no machine has been made,” Mashiko says “Applications-wise, we are able to show the utility of diffraction casting in both well-established and emerging areas of computing that move beyond classical operations — the 16 fundamental logic operations at the heart of most information processing right now, as well as a novel functional system: quantum.
