A new computational approach that breaks classical HPC and quantum limitations

There is much anticipation about the possibilities of exascale supercomputing and quantum computing, but there is an emerging paradigm of advanced computing that transcends the limitations of traditional high-performance computing - and is much more practical than quantum computing - when it comes to complex multivariate problems.

This computing technology, known as laser-based processing units (LPUs), has the potential to leapfrog to the lead in solving the toughest optimization challenges.

Before we get into how LPU technology works, let's quickly discuss the problem at hand. Industries such as finance, supply chain/logistics, and manufacturing need new solutions to address complex multivariate challenges ranging from optimizing financial portfolios and transactions to improving decisions related to vehicle routing and warehouse optimization, as well as scheduling and assembly line balancing.

This requires finding the best result from a set of possible solutions that satisfy certain constraints ("constrained optimization"). For NP-hard optimization challenges, the time required to find a solution grows exponentially with the size of the problem, which is determined by the number of variables or constraints: hard optimization problems all have one thing in common: solving them requires powerful computational capabilities.

Today, many optimization problems are solved by classical computers using approximations (e.g., heuristic algorithms). Unfortunately, these solutions are often of low quality and the time to solve the problem increases with the size of the problem. In order to solve the problem, a large number of parameters are usually omitted in order to solve it in a reasonable computation time, leading to uncertain results. Even the most powerful supercomputers with a performance of more than ten trillion FLOPS may reach their computational limits.

As for quantum computers, they are still immature, and the day when they will be affordable and scalable is a long way off. There are many quantum modes available, each of which faces engineering challenges: for example, the need for ultrahigh vacuum, specialized components, and ultracold and complex stabilization systems. Despite these extreme conditions, quantum bits continue to interact with their surroundings, making them error-prone and thus less reliable and accurate.

Quantum annealers have been commercialized. However, these room-sized devices cost up to $15 million and require the same extreme operating conditions as quantum computers. Moreover, because of their low connectivity, they cannot be easily scaled to solve challenging real-world problems.

In May 2023, Israeli startup LightSolver unveiled the first pure Laser Processing Unit (LPU): a quantum-inspired solution that employs all-optical coupled lasers, requires no electronics for computation, and makes it as small as a traditional desktop computer, while offering unmatched scalability, low power requirements, and room temperature operation.

LightSolver LPU

The Laser Processing Unit is a new computing paradigm that utilizes lasers for computation. It offers several advantages over traditional computing methods, including greater speed and accuracy. the LPU's small size allows it to be built with commercially available components, making it easier to bring to the mass market.

The LPU uses all optically coupled lasers and requires no electronics to perform calculations, making it as small as a conventional desktop computer, while also featuring scalability, low power requirements, and room temperature operation. It utilizes the laser's ability to interfere and thus satisfy all the constraints of a given problem.

First, users of the LPU are asked to formulate the quadratic unconstrained binary optimization (QUBO) problem - a mathematical problem that seeks to find the best combination of variables that satisfies specific constraints. The conditions of the problem are then mapped to "obstacles" in the optical path in which the laser operates.

Due to the wave nature of the laser and the proprietary mapping, the laser beam converges to the desired solution. The solution is then measured and converted into the user's business language. With this approach, scientists can solve optimization problems at the speed of light, orders of magnitude faster than other current techniques.

This approach encodes the constraints of the optimization problem in the relative phases ("laser bits") of the lasers, and interacts the phase states by using arrays of tightly coupled lasers that diffract the light from each laser to all the other lasers in a controlled manner. This design allows for complete connectivity between all lasers, pairwise full-spin interactions on a desktop-sized device, operation at room temperature, and requires only a small amount of energy.

However, unlike quantum, the device is portable, not subject to environmental sensitivities, and does not require error-correction protocols. In addition, laser-based supercomputers can be scaled significantly without increasing the size of the device. In an LPU, each variable is a laser spin; this also differs from a quantum computer in that each logical spin is also a physical spin, so it is not affected by overhead spins that need to be allocated.

A recent experiment at LightSolver demonstrated that lasers can compute puzzles without the need for electronic devices (currently in the final stages of peer review).

Forty laser bit processors performed the computations and showed promising results in solving complex optimization problems, with a two-order-of-magnitude speedup compared to powerful GPUs, initially demonstrating the enormous computational potential of LPUs. The experimental team said, "The new LPU paradigm, which utilizes existing mature laser technology, promises to surpass quantum computing and supercomputing in the coming years."

Reference link:

[1]https://thequantuminsider.com/2023/05/09/lightsolver-announces-laser-based-processing-unit/

[2]https://insidehpc.com/2023/07/overcoming-the-limitations-of-classical-hpc-and-quantum-a-new-computational-method/

[3]https://lightsolver.com/

[4]https://arxiv.org/pdf/2209.03788.pdf