The new building, which houses the largest quantum computer in Europe, doesn’t exactly look impressive. If it had windows in its wood-clad outer wall, it could also be the day-care center of Forschungszentrum Jülich, where it stands between gray institute buildings. Kristel Michielsen unlocks the only outside door that leads into a foyer. There are two glass doors on the back wall. Orange shimmers through the one on the left, as if from a bar. “The D-Wave computer works in there,” says Michielsen, head of the Quantum Information Processing group. She opens the door, inside a humming welcomes the visitors and a black box the size of a bathroom. Orange letters glow along one edge: “Advantage” they proclaim – “advantage”.
The name of the quantum computer is a promise: The computer from the Canadian company D-Wave Tasks is supposed to solve tasks much faster and more precisely than even supercomputers can do today. However, it is not the first of its kind: since 2020, two more copies of Advantage have been working at the D-Wave headquarters near Vancouver. Researchers and companies share computing time there via cloud access. So far, Michielsen’s Jülich team has done the same. “But now we can try Advantage on our own together with the industry,” says Michielsen. “We want to explore the potential of this technology.”
Business is interested in Advantage because of the kind of problems its creator designed it for: optimization problems. That sounds scientific, but it is everyday practice. Organizations often waste time, human resources and money by choosing from a plethora of possible solutions far from the best possible one. Even with supposedly simple tasks, the optimum is hidden in a thicket of complexity, as Michielsen’s colleague Carlos Gonzalez Calaza illustrates.
Optimizing a field with just a few plant varieties can overwhelm classic computers
“I plant vegetables on my terrace,” Calaza begins his remarks in his office near the new building with the quantum computer. He wanted to grow several types of plants in the small space. There are plants that like to grow next to each other, and there are those that don’t like each other, he explains. Tomatoes and cucumbers, for example, should not be in two adjacent pots, but tomatoes and lettuce should.
As Calaza tried to optimize the arrangement of his plants with a pad and pencil, he realized how complex the task was. Even for three varieties there are six ways to string them together. The number of combinations multiplies with each new variety. If you distribute the pots on an area, you have to consider even more neighborhoods. “If you only change one pair, it affects the whole field,” explains Calaza. A dense network of possibilities emerges, in which there are a number of solutions that are so-so. But where in the thicket is the very best solution hidden, one in which every plant has the ideal neighborhood?
The task belongs to a class of problems whose complexity overwhelms even the most powerful classical computers. So Calaza figured this would be a good way to test D-Wave’s calculator. This is more than gimmick. Because there are plenty of tasks whose complexity quickly explodes in everyday business life. Airports have to process hundreds of planes at a few gates every day as smoothly as possible. The optimum is difficult to find because of the many possibilities and a number of boundary conditions. Some aircraft are too big for a certain gate or block neighboring gates.
The logistics of supermarket chains is also highly complex, the production process in factories, the balancing of risks and returns in stock portfolios. Even machine learning, such as that used to control autonomous vehicles, is at its core an optimization task.
A few hundred qubits can hold more values than there are atoms in the universe
Quantum computers could crack these tough nuts. Your greatest strength: You can process almost any number of options in parallel. The basis for this is the “quantum bit”, or qubit for short. It is an extension of the bit, the smallest unit of information in a classic computer, which can switch between the values 0 and 1. Quantum physics ignores this either/or. Particles such as atoms or electrons can assume several states simultaneously, for example be in two places at the same time.
If you define one state as 0 and the other as 1, you get an information unit that can record both values simultaneously. Each additional qubit doubles the capacity. Two qubits hold four values in parallel, three qubits hold eight values, and so on. A few hundred qubits can hold more values than there are atoms in the universe. Physicists use lasers or microwaves to manipulate and interact with qubits. This is how they execute algorithms, i.e. the qubits calculate simultaneously with all stored values.
Advantage is a special kind of quantum computer and not to be confused with those from IBM or Google, which are not yet practical. The developers of these tech giants have a different claim: the computers should be freely programmable and accelerate a wide range of applications. With more than 5000 qubits, Advantage has significantly more than these quantum computers, but remains limited to optimization tasks.
He always solves these with the same method. The 5000 qubits represent all possible combinations of the problem at the same time. In the garden example, these are all possible arrangements of the plants in the pots, both favorable and unfavorable. The algorithm simultaneously calculates a “price” for all possibilities, which is higher the more unwelcome neighbors a plant has. Physically, the price is reflected as energy. Thus, the optimum has the smallest energy. Because physical systems tend towards the energy minimum, the suboptimal solutions disappear from the qubits by themselves. The optimal solution remains and can be read out.
“We were able to increase the return on investment by ten percent in just a tenth of the time.”
The qubits are made of the metal niobium, which conducts electricity without resistance at a few degrees above absolute zero temperature of minus 273 degrees Celsius. In this state, the current obeys quantum physics. If you form the metal into a loop, the current circulates left and right simultaneously. The two simultaneous current directions form a qubit. Advantage’s qubits fit on a fingernail-sized chip. The refrigeration technology, the insulation and the control electronics take up a large part of the bathroom-sized box in Jülich. However, the visitor is not allowed to look inside.
What is more important anyway is what the black box spits out. Users of the sister computers in Canada were impressed at the inauguration of the Jülich computer. One of them is Gonzalo Gortazar, head of Spain’s Caixabank. His house optimized an investment portfolio using the D-Wave calculator. “We were able to increase the return on investment by ten percent in just a tenth of the time,” says Gortazar.
Vern Brownell, Managing Director of D-Wave-Systems, describes an example from logistics. A Canadian supermarket chain asked because the food supply chains had become more complex in the corona pandemic. “It took them about 25 hours per site per week to calculate an optimal solution,” says Brownell. With the help of D-Wave’s computer, the problem was solved in two minutes each time.
The quantum computer from Canada is therefore optimizing at high speed. But it remains open: does it really outperform classic computers? So far, researchers have not found a clear answer. First there is the question of how reliably Advantage finds the optimum and not just the second or third best solution. To do this, Kristel Michielsen’s team gave the computer problems whose optimal solutions it already knew. “We saw very different results,” reports Michielsen. In some cases, the quantum computer has found the optimal solution, but in other cases not the very best. Sometimes rather bad solutions.
Researchers are also back and forth on the question of whether Advantage comes up with solutions faster than the fastest previous optimization methods. With tricks, classic computers also find very good solutions in the thicket of complexity, even if they are not the very best.
At first it looked like D-Wave would have the edge. The company published a research paper in 2017, according to which the D-Wave computer is 2500 times faster than the best classical methods. But the criticism followed promptly: the classic methods used could still be greatly simplified, wrote researchers led by Salvatore Mandrà from NASA’s Ames Research Center. It is possible that D-Wave’s customers had not exhausted the conventional methods either. “So far, there has been a lack of scientific evidence that D-Wave computers have any advantage,” says quantum computer expert Scott Aaronson of the University of Texas at Austin.
“An honest comparison is very difficult,” says Kristel Michielsen. She views Advantage as a research project. With the quantum computer on her own campus, her team wants to better understand the physics of qubits. “Perhaps we will discover new interactions between the qubits that could bring an advantage,” says the quantum computer scientist. At the same time, the practical potential of Advantage is to be researched in cooperation with companies. “Projects are planned with the auto industry or with energy supply companies,” says Michielsen. The largest quantum computer in Europe will therefore have ample opportunity to show off its skills under the scrutiny of many researchers.