When it comes to developing smaller and smaller microchips to fit in our smartphones, wearables, and other mobile gadgets, experts believe that we’re getting closer and closer to a limit dictated by Moore’s Law. However, researchers at Hewlett Packard Labs (now a part of Hewlett Packard Enterprise) have developed better ways to optimize computing using light instead of electrons.
Research focused on developing optical chips isn’t entirely new. What the team at HP Labs has created, however, is supposedly one of the most complex optical chips around and can perform optimization tasks more efficiently than existing hardware. “We believe that it is [the biggest and most complex] by a wide margin,” said team member Dave Kielpinski. The research is featured in the IEEE Spectrum.
Developed through the U.S. Defense Advanced Research Projects Agency’s Mesodynamic Architectures program, the new device features 1,052 optical components that crunch numbers together. This new device is a light-based take on the Ising machine.
Simply put, an Ising machine is designed to solve through problems involving large numbers of competing alternatives by finding the best solution. Ising machines encode problems as temperature fluctuations, the solutions for which are determined by how the spin of electrons respond to heat change and settle over time.
With the device HP Labs developed, chips use light beams instead of electrons. Basically, these light beams are polarized to imitate the two kinds of electron spin. The light moves around the chips with small heaters to encode the problem and stops only when the light reaches a steady state that provides the answer.
Certainly, this device pushes the limits of photonic design, according to Dave Kielpinski, which is one of the project’s key aims. “One of the things we’re proudest of is our computer-aided layout tools,” he said.
This optical device is capable of optimization process faster than its rivals. For example, it can breeze through the classic traveling salesman problem — a computational problem that looks for the most efficient route between a number of points — better and faster than existing chip hardware or even other optical chips, according to the HP Labs team.
In sum, the main advantage of such a technology is optimized optimization — faster information processing and larger-scale number crunching. On top of that, light-based chips are also capable of using less energy.