Cloud Hosting

Many experts believe that once quantum computers are big enough and reliable enough to solve useful problems, the most common deployment architecture will be to have them serve as accelerators for supercomputers. Several algorithms have already been developed to run on this architecture, including the most famous one Shors algorithm, which will one day crack many of todays public key cryptosystems.

To experiment with this arrangement in Finland, the Finnish Technical Research Center (VTT) cooperated with CSC, who run LUMI, Europes fastest supercomputer. VTT and CSC have been connecting progressively larger quantum computers to LUMI to allow users to experiment with algorithms that will make the best use of both types of computers. The classical computer stores most of the data and executes most of the steps in the algorithm, handing off other steps to the quantum computer and then reading the results.

VTTs first quantum computer was completed in 2021 and used 5 qubits. They then upgraded to20 qubits last yearand are aiming for 50 qubits by the end of 2024. Universities and research organisations have already started using the hybrid service to solve trivial problems.

This is sort of a rehearsal for the future, saysVilleKotovirta, leader of the VTTteam that develops quantum algorithms and software. It allows people who develop supercomputing algorithms to start thinking about what they will be able to do when a bigger quantum computer works alongside a classical computer. They can use the existing service to practice writing algorithms.

We were the first to set up the hybrid configuration in Europe, but others are following, says Kotovirta. In June 2023, the European HPC joint undertaking acceptedsix other projectsto build similar architectures. These will be in Czechia, France, Germany, Italy, Poland and Spain.

Kotovirta is responsible for research in quantum algorithms and development of software to enable access to VTTs quantum computing. Part of his job is to hire new talent, which he says is not easy. Some of the graduating students already in Finland are interested in quantum computing, but they dont have real world experience.

The people who have work experience in Finland already have a job somewhere else in the ecosystem, and people from outside Finland are hesitant to move to a cold climate. Having said that, some of the outsiders are impressed enough with theFinnish ecosystemto overcome concerns they may have about the weather.

Were all learning because its a new field and its changing all the time, says Kotovirta. There are new inventions, new platforms, new devices, new algorithms and new ways of programming. To keep up, we try to hire mathematicians, physicists and computer scientists.

Kotovirtas team is developing several types of algorithms for hybrid architectures. One is a set of optimisation problems, called quadradic unconstrained binary optimisation (QUBO), which can be solved using quantum annealing or quantum approximate optimisation algorithms (QAOAs).

We have built quantum algorithms for analysing graph data and identifying the community structure of networks, he says. The data comes from complex networks, like technological or biological networks systems.

The team is also developing algorithms for quantum chemistry, with focus on reducing the complexity of amolecular Hamiltonianto improve on simulations of molecules. Similarly, they are working on synthetic biology, where they generate new proteins, with certain desirable features.

Another area of focus is quantum machine learning especially quantum generative machine learning, models that learn from existing data to produce novel samples.

Most people have heard of generative AI in the context of fake AI, where it is used to create images, text and sound, says Kotovirta. These same techniques can be applied to science, learning from something that already exists to create something new. We are finding ways of improving these techniques with quantum computing to generate new proteins.

The most difficult part is proving that quantum machines have benefits over their classical counterparts, says Kotovirta. Current quantum computers are real computers, and they can do real calculations and solve real problems.

In that sense, they are already doing useful things, but the problem is that sizes are very small, because the current systems are inefficient in comparison to their classical counterparts. In order for quantum computers to solve something more efficiently than classical computers, fidelities need to improve.

Fidelity refers to the success rate of a single operation on a quantum computer. The higher the fidelities, the better success rate of the overall computation. We are currently in an era of noisy intermediate-scale quantum (NISQ) devices, which have already shown that quantum devices can simulate things classical computers struggle to simulate. However, so far, the results are so trivial that they dont solve real problems.

As fidelities improve, were approaching the era of utility-scale algorithms, utility-scale quantum computing, says Kotovirta. This will happen when the fidelities are good enough to run certain algorithms that are tailored for the topology of the device youre using. That will give us results that classical computers cannot replicate, but only for very specific use cases.

These algorithms could be used to simulate quantum systems related to material sciences or chemistry, for example. Although you cant claim general quantum advantage, for those specific use cases you can demonstrate advantage. Finlands strategy is to make a difference on the world stage through the use cases in which quantum advantage can be achieved in the not-so-distant future.

It isnt easy for smaller countries to compete with larger economies. However, it is possible for them to find a niche that allows them to actively contribute on a global scale in one or more specific areas. In that regard, we are doing the right things, he says. Hopefully, we will continue to do so in the future.

Read more:
Finland develops quantum algorithms for the future - ComputerWeekly.com