A New Optical Chip Makes Light Work of Optimization Problems

As Moore’s Law stutters, computing companies are looking to develop ways to calculate using light rather than electrons. Now researchers at Hewlett Packard Labs have built one of the most complex optical chips yet, and claim that it could be used to perform optimization tasks like the traveling salesman problem more efficiently than regular hardware.

IEEE Spectrum reports that the team has built a device that features 1,052 optical components that all work together to crunch numbers. The device is a light-based take on a so-called Ising machine. Such devices encode a problem as temperature fluctuations, and determine a solution by identifying how the spin of electrons settle over time in response to the changing heat.

The Hewlett Packard Labs chip trades electrons for light beams, which it polarizes to mimic the two types of electron spin. The light churns around the chip with small heaters encoding the problem until the light, too, reaches a steady state that provides an answer. IEEE Spectrum neatly describes the technicalities of how this works. The researchers claim that the approach can be used to identify solutions to complex optimization problems faster than normal chips.

Other optical computing techniques promise similar advantages. As we recently reported, laser-based computing approaches are being used to analyze genetic data and intelligently compress information faster than normal computers. Along with their speed gains, light-based chips could also use less energy.

Such advantages explain why chip giants like Intel are also working out how to build optical computing hardware. The Hewlett Packard Labs creation is another milepost on the road to making lighter work of the problem.

(Read more: IEEE Spectrum, “Intel Tries to Rearchitect the Computer—and Itself,” “Light Chips Could Mean More Energy-Efficient Data Centers,” “Computing with Lasers Could Power Up Genomics and AI”)