快猫短视频

A machine made from a 5-kilometre-long fibre optic cable coiled into a box a few metres across could give quantum computers a run for their money on certain tasks. The device, which performs calculations using pulses of light, will be one of the most sophisticated optical computers ever built.

The device, which is being built by Hiroyuki Tamura and colleagues at Japanese tech firm NTT, is specially designed to solve optimisation problems. These involve finding the best solution out of many possible ones, such as the best way to balance an energy grid or the most efficient way to schedule deliveries around a city.

Rather than relying on electrical circuits, the device will shoot 100,000 optical pulses into the fibre optic coil where they will follow each other round in a procession. Optimisation problems typically involve complex interconnected networks, so a special computer chip encodes the characteristics of this network into the pulses as they fly past.

Each pulse is given a positive or negative value and told which other pulses it should interact with. The pulses then circulate around the coil tens or hundreds of times, interacting with each other until they reach a stable state representing the best solution.

The team has previously built a version that could squeeze 2000 pulses into a 1-kilometre-long fibre. Now they are building one that fits 100,000 pulses into 5 kilometres of fibre. That will allow the computer to model optimisation problems with up to 100,000 moving parts, which could start to tackle complex real-world problems like optimising the layout of communication networks.

鈥淭his would definitely be quite groundbreaking,鈥� says Charles Roques-Carmes, who studies optical computing at the Massachusetts Institute of Technology. At that scale, conventional computers would take centuries to find exact solutions and even programs designed to find approximate solutions to real world problems would be slow, he says.

Pulses vs qubits

Optimisation problems are challenging for conventional computers because the number of possible solutions rises exponentially as the problem grows, requiring ever more computing power. Quantum computers have a natural advantage on these problems over conventional computers, but the technology is still in its infancy.

In a , NTT鈥檚 computer containing a 1-kilometre-long fibre was tested against D-Wave, the first commercial quantum computer, on a series of experimental optimisation problems. While the D-Wave did better at small problems, NTT鈥檚 device did significantly better on bigger problems with many interconnections, which are more indicative of real-world problems.

The two machines work in very similar ways, but D-Wave encodes problems using superconducting qubits 鈥� quantum equivalents to the bits in a computer 鈥� rather than optical pulses.

The NTT computer can connect every pulse to all the others, whereas D-Wave can only link a limited number of its qubits. That connectivity makes it possible for the NTT device to solve larger, more complex problems, says Tamura.

More powerful and broadly applicable quantum computers being developed by companies like Google and IBM may ultimately outperform LASOLV, says Tamura, but it will probably be decades until they are large enough. The 100,000 pulse machine should be ready within a year or two, he says.

However, there are still technical challenges to be solved, says Tamura, namely coping with more and faster pulses and making them more resistant to environmental disturbances on their 5-kilometre journey.