Quanta, qubits and combinatorics

While some aspects might sound like they've jumped off the page of a science fiction book, others are already market-ready. I discussed this topic with Anne-Marie Tumescheit, one of maybe 10,000 people in Germany who know something about quantum computing. Ms Tumescheit, you work in Business Development for the Japanese tech company Fujitsu in Berlin. Does Fujitsu already use quantum computers to solve problems?

Firstly, I'd like to say how delighted I am to be here. And no, we don’t yet use any actual quantum computers to solve problems. But we are already working with quantum-inspired technologies in order to help our customers on certain issues.

In terms of our conversation, it probably helps that you not only studied bioinformatics but also worked in scientific communications at the Fraunhofer Institute, Germany’s biggest research network. As a result, you have successfully made highly complex, technology-related topics easy to understand. Can you translate quantum computing for us?

I will try my best to put it in straightforward terms. Having said that, precise scientific details may be glossed over at certain points as a result. But it’s important to get an introduction to the terminology and realise what quantum computers might one day entail for our world.

On that note, let’s get started with the basic vocabulary. What is a quantum?

That seems like the best place to start. A quantum is the smallest possible amount of energy. For a start, it is a unit of measurement, just like centimetres or kilograms. A quantum also describes the smallest possible packet of an electromagnetic wave, such as light, radio waves, microwaves and other similar waves. This smallest packet cannot be broken down any further. It’s a bit like having a bag of chips with ketchup, with the smallest portion being one chip.

In addition to a quantum, another basic term is used in quantum computing. Instead of the bits we encounter in standard computing, the qubit is the smallest possible unit of information that can be processed by a quantum computer. It sounds like it might be pretty complicated. Could you explain it for us in simple terms?

As you've pointed out, we’re familiar with the term 'bit' from the world of computing. You can process bits and they have two possible statuses: one or zero. In other words: power on, power off. Simple as that. The fact that these bits are so well interconnected means that any program or calculation can be run on the computer.

With a qubit, we have the advantage that its state does not have to be just zero or one, but can also be any state in between. This principle is known as superposition. That is to say, two qubits with different states can be superimposed. You can imagine that like little balls stacked on top of one another. They can also represent many more states. In theory, a limitless number of states.

With quantums and qubits, you have already explained a few basics of quantum computing to us. How do these kinds of computers differ from the ones you would find in a computer centre?

In general, it's important to note that a quantum computer offers completely different options than a standard computer and, as a result, works completely differently. The fact that a qubit can assume many different states, each of which is associated with a certain probability, means we now have the opportunity to save much more information in this cube.

There’s another characteristic that also comes from quantum mechanics: They are entangled. This means that qubits can exchange information between each other without losing any time. As a result, if I change an image, I also change the entangled qubits at the same time. This means I create a kind of chain reaction, like dominoes, but no time is lost during this reaction. A standard computer is simply not capable of this.

Do quantum computers all work in the same way, or are there fundamental differences?

There are two fundamentally different ways in which quantum computers can work. On the one hand, we have quantum gates. Gates have most in common with our standard computers because they are universal and can solve all kinds of problems. Unfortunately, research on gates remains in its infancy. The second variation is the quantum annealer, which is best suited to solving optimisation issues.

This second variation, annealing means something hot cooling down over time. Can you explain what gets annealed?

This method that has actually been around in mathematics for a long time. It describes translating a problem into an energy equation. This applies in particular to optimisations, i.e. problems that are generally deductive. The best option needs to be found among an unimaginably vast number of options.

By contrast, editing translates precisely this optimisation problem into an energy equation. You can envisage the result like a mountain range. We have mountains, the maximum values, and valleys, the minimum values. As a rule, I either want to find the highest mountain or the deepest valley. And that is described by the energy equation used to form this mountain range. At the end of the day, a quantum annealer can determine the highest mountain or the deepest valley.

What kinds of tasks can this mountain landscape produced by the quantum computer help with?

These mountain landscapes are, as I mentioned, the solutions of energy equations. They can be particularly effective at addressing optimisation issues but they can also be modified to conduct searches in particularly large quantities of information. We are already capable of solving these kinds of problems in this way.

Using the theoretical super-performance of quantum computers still sounds highly complex and only possible to a very limited degree in practice. Why is that?

Well, quantum computers work with very tiny particles. Namely atoms, or smaller particles, such as electrons, protons and neutrons. We know that atoms have the great property of always being on the move. In lay terms, this is known as heat. But heat is nothing more than a certain amount of kinetic energy in the particles.

If these particles are always flying around, we can’t always rely on them to act in a certain way. It’s a rather complex endeavour. So for that reason, it’s important we cool them down to almost –273 °C. Or, to put it another way, zero degrees Kelvin. We need to cool them to this kind of temperature so that they stay still. Then we can catch them and push them into a particular pattern so we can work with them. Otherwise, they just shoot off in all directions.

These kinds of computers also require an extreme kind of shielding, don’t they?

That’s right. That’s also because they work using such tiny particles. And they are influenced by everything around them – no matter whether that’s cosmic rays, radio waves that can be found anywhere on Earth, or even if an employee just walks past with their mobile phone switched on. All of these energy sources influence the particles because they are extremely small and therefore very easy to influence.

Those sound like some major obstacles to the use of quantum computers in practice. How long will it be before real quantum computers are economically viable?

Well, as with most matters, estimates vary. Scientists are yet to come to an agreement. Some say five years and some say ten. And of course anywhere in between, and beyond. Personally, I think that we might be able to imagine them being economically viable in ten years’ time. We are bound to make significant progress in the next five years in terms of the size of quantum computers. And in five years, we will be able to start working with sensible sizes. But it will still take around ten years until we can actually reach economic viability.

So, the ones in commercial use today are not actually quantum computers, but quantum-inspired bridging technologies. What does that mean?

There are two types of quantum-inspired bridging technologies: Firstly, the ability to simulate quantum effects, just like lots of things can be simulated on a computer, such as migration flows or growth. Of course, at the moment, the topic on everyone's lips is the course of the pandemic and spread of infections. All of these calculations are simulations based on physical equations. The same thing can be performed with quantum effects. But there, too, we’re talking about a very small area of application, where we can simulate 20 to 30 qubits.

When we speak of emulators we refer to very special hardware that is inspired by quantum effects. For example, our company produces a digital annealer. Microsoft now also has something on the market. These devices, digital annealers, make it possible to compute much, much bigger numbers. We’re talking 8,000 to 100,000 qubits – which means it is already possible to solve problems on an industrial scale.

Where can this technology currently be found in commercial use?

We worked with BMW, for example, to optimise the movements of robot arms. Today’s vehicles aren’t painted by hand any more, but by robots. These robot arms have several joints, allowing them to move in a wide variety of ways, in every possible direction. Two things are important here: First of all, they can’t get in each other’s way or damage the car. And secondly, they should of course move in an optimal way to paint the car as quickly as possible. This is one example of deductive optimisation, where you simply deduce all possible movement patterns of a single arm in order to find out the shortest time and safest pattern of movement for these robots.

A second example is, in fact, partly inspired by the pandemic. We all know the rules: For example, a maximum of two households can sit together at a distance of 1.5 metres, things like that. We applied this to a stadium – which enabled us to optimise the stadium’s capacity while sticking to the permitted number of people and social distancing rules. This allowed a much higher capacity utilisation of the stadium than if I had worked it out by hand because I would have to optimise so many different factors. Always with the same conditions: Who’s allowed to sit with whom? Who has to go which way, and who is allowed to meet whom? And these options can be organised in a way that enables me to safely get as many people as possible into the stadium.

Now let's consider the extent to which quantum computing bridging technologies are already able to help in the port and logistics sectors, which are of course HHLA’s prime concern. HHLA is part of a research consortium along with the Institute for Laser Physics at University of Hamburg and other partners that aims to look into how quantum computing could help optimise shipping routes or supply chains, for example. By doing so, we aim to save energy costs and help to protect the environment. Is that a promising approach in your opinion?

Yes, generally speaking, I think you could say so. Let’s take the optimisation of global supply chains, for example, in today’s world, where there are lots of stop-off points where things are unloaded and stored. This is a huge deductive optimisation process and a quantum computer would be very helpful there.

Protecting the climate is also often based on simulations that account for an extremely large number of factors. Take this rather daunting idea: If a butterfly flaps its wings in China, we can feel the effects here in Europe. Of course, we don’t notice it directly but the climate is actually influenced to this kind of degree of detail. The same goes for the weather – and these are things that are based on a great deal of calculations.

So, in order to play to its strengths, a computer of this kind would need to account for as many variables and unknowns as possible.

Exactly. It is crucial that we have lots of options, can create lots of combinations and then also have a lot of data to process.

What would you also consider a good, logical example of the use of a quantum computer at a port?

We are actually working with the HPA on a project in the Port of Hamburg that deals with optimising traffic flow. The optimisation of traffic is, in general, a good example of the use of quantum-inspired technologies, and, in the future, actual quantum computers.

There are already a few solutions for managing backlogs at the port, including intelligent traffic-light congestion management. Why is the high processing power of standard control systems not enough in this case?

The problem with adaptive traffic light systems is, unfortunately, that each junction is self-contained. This means that if a large convey approaches a junction controlled by three traffic lights further on, or if a certain direction of travel is particularly busy, the junction will only become aware of this once the cars have already reached it. This leads to delays and can cause traffic jams, which we want to prevent.

The ideal solution, therefore, is to give all traffic lights the option of communicating with one another. As follows, the traffic lights aren’t able to figure things out among themselves but we can instead include all traffic lights in a system that monitors current traffic flows and simply calculates the optimum combination of traffic light sequences. And we can implement this across all the lights, making the first traffic light change in a way that optimises the traffic flow, even at the third, fourth or seventh set of lights.

One related issue with relevance to the port would be: How can routes using intermodal traffic – i.e. water, rail, road – be optimised in order to move containers from A to B more quickly The popular example of the travelling salesman explores the same issue. Our salesman needs to travel to so many cities. But what’s the fastest route? Can Fujitsu help out in terms of port logistics with its digital annealer?

Yes, this is something we have actually already done. Our earliest example is the optimisation of the Japanese postal service, where we were able to optimise the distribution of parcels, or in the US, where we were able to optimise the distribution of medical supplies such as PPE, masks and those kinds of things. They had to be supplied from a central warehouse to all sorts of places. On the one hand, our solution contributes to cost efficiency. On the other, it gives people peace of mind that things can be supplied on a just-in-time basis, no matter whether it’s production goods or medical staff.

In terms of freight traffic, Deutsche Bahn has a massive network to oversee. In this case, it wanted to optimise the assembly of its trains and keep a flexible number of goods wagons available. Fujitsu once again stepped in as a solution-oriented partner. Can you tell us a bit about both the problem and the solution?

When it came to the project with Deutsche Bahn, the idea was to determine or optimise where and how Deutsche Bahn sent its goods wagons and freight trains through the country, and where they were held in reserve. Generally speaking, goods wagons are booked by companies a year in advance. Needless to say, unforeseen circumstances can mean that companies actually need more or fewer than originally anticipated. In order to solve this problem, we looked at the whole picture using the annealer and our algorithms and found a way to optimise this process.

This approach could also be an option for HHLA’s rail subsidiaries, for example, in order to optimise rail traffic or the transportation of goods. However, each optimisation requires a certain magnitude. If you only have five options for optimising something, I can do that in my head and don’t need a computer – neither quantum nor standard.

The key IT systems in the port are networked and are constantly exchange data with one another. But if your company is to help optimise systems as complex as global supply chains, are there suitable interfaces for collecting all the data required and processing it in a consistent way?

Data formats and data interfaces are a prime topic. Globally, we have a whole smorgasbord of potential data at our fingertips and we know how it is structured, what it looks like and how complete it is. Some companies have enormous data sets that are very complete, while others only have fractions of this amount of data.

So, before we can even use it, we need to prepare data in a solution-oriented way. We need to look at what we want to achieve and create the relevant interfaces or cleanse the data beforehand, which is another very broad field of work involving data scientists.

I can now imagine that you may have interested HHLA or other companies in the port involved in logistics in partnering up to look at certain issues. What does a partnership with your company entail? You don’t just sell a computer with a patented quantum chip and then say: Well, off you go, then!

Correct – we don’t do that. Every partnership starts with a consultation phase. Together, we look at the problems faced by the company and which of these problems could potentially be solved by the digital annealer. The individual problems are then categorised. We always look at where there is room for improvement.

As I already mentioned, you need a certain order of magnitude in order to make it worthwhile. The problem is then computed in a proof of concept. We develop the necessary algorithms needed and discuss areas that can be optimised with the company. We also look at the real-life requirements, whether our solution is necessary, whether they need it and how they might be able to use it.

Ms Tumescheit, you have succeeded where my old physics teacher failed – there is no way he could have known anything about this subject. You have made me feel like I have understood at least the basic principles behind quantum computing as a future technology, and its potential applications. Thank you for the fascinating and informative conversation!

Updated on April 19th, 2022.

The interview was conducted by Oliver Driesen.

Interesting information or suggestions taken away? Don't forget to subscribe to our newsletter HHLA Update and stay tuned.

https://hhla.de/en/update