It swallows up tens of billions in research, but is not functional, the experts interviewed candidly admit. It might make our old computers outdated, but it couldn’t handle our tax return. We don’t really know how to program it, we’re not even sure that it will work one day, but the big companies are competing with each other with triumphant announcements.
And yet, no, the quantum computer in 2023 is not a huge hoax.
How is it possible ? Difficult for ordinary mortals to understand that IBM can present in Bromont its supercomputer with “overdrive” computing capacity, which Google claims to have demonstrated in 2019 the superiority of the quantum computer… with machines that we do not estimate not functional.
In the quantum universe, we are not close to a paradox. “We must first define what we mean by “functional,” immediately warns Alexandre Blais, scientific director at the Quantum Institute of the University of Sherbrooke. If the question is: “Does it work, are there people using it to make useful calculations that are of interest to society?”, the answer is clear. It’s no. »
But he immediately specifies: “It is used as a research tool, as a teaching tool, as a learning tool by future users of a machine. »
Gaël Humbert offers an analogy. “We have planes that fly for a few seconds, we know that we are going to make them fly further and further, but we still don’t have a pilot. » The director of business development at the Digital and Quantum Innovation Platform (PINQ2), a non-profit organization set up by Quebec in 2020, laughs when we ask him naively if we could file our income tax return with one of the quantum supercomputers in the world.
“Why should he file taxes? There are daily tasks that are today carried out by traditional computers. […] It’s like asking a plane to deliver a pint of milk. »
In fact, adds Jean-François Barsoum, executive director of Innovation at IBM, we are at the point that marked the takeoff of computing in the mid-20th century. “We didn’t know what classical computers were going to be used for. People thought there would be a global market of five computers. »
The design of a quantum computer faces three main difficulties in 2023 (see the “Obstacles” capsules). The most optimistic hope to have a reliable computer within three years, but the most commonly accepted horizon is around ten years. For example, IBM Quantum System One in Bromont has 127 fundamental calculation units called “qubits”. Some scientists believe that it would take a million for a computer to be effective, an assessment that is not unanimous.
Which does not prevent researchers from now trying to understand its usefulness. The most obvious avenues are the already known weaknesses of classical computers, which are gigantic supercomputers trying to find a single path to a single result. The quantum computer can theoretically take an exponentially greater number of paths to deliver a result.
Calculations that we believe to be simple are today technically impossible or very difficult to solve with a conventional computer, however powerful it may be.
This ability to juggle prime numbers makes the quantum computer potentially superior, by far, for cryptography.
The other well-known example is the so-called “traveling salesman”. He must visit at least once a certain number of cities whose distance he knows in pairs. Finding the shortest path to 20 cities would take a typical computer 964 years. A quantum computer could do it in a fraction of a second. This is what makes it a significantly more effective tool in areas such as the optimization of logistics chains, transport, digital modeling to discover new pharmaceutical molecules, explains Thierry Souche, vice-president of technology at the French company. OVHcloud cloud computing. Everything happens, in fact, as if the quantum computer managed to reproduce with its qubits the behavior of a system with complex interactions. “Quantum is analog,” he explains. I take a system, I excite it, I let it fall and I see how it is. It is governed by physical properties which are roughly the same as those I want to solve for. »
The experts consulted are unanimous: recent years have demonstrated that the quantum computer is not a chimera and that its advent will shake up society, in combination with classical computing which will retain its usefulness. The idea in 2023 is to not be undressed when quantum computers are fully functional. “Take the example of cryptography: we don’t know if we will be able to decode it,” emphasizes Gaël Humbert. But imagine the impact on businesses if we succeed. We have to be able to identify risk areas, otherwise it will hit us hard. »
We must avoid reproducing with quantum the current situation of artificial intelligence, while breakthroughs in academia struggle to be applied in industry, warns Jean-François Barsoum. “The idea is to create an ecosystem where researchers work with companies on applications. »
In a classical computer, information is encoded in bits, which can take the values of 0 or 1. The quantum computer is made up of “quantum bits” or qubits, the understanding of which requires a minimum of physics. quantum. These qubits do not take a defined value of 0 or 1, rather they refer to a set of probabilities between 0, 1 and intermediate stages. Mathematically, we consider that they have all these values until we measure them. This is called layering. The other applied quantum phenomenon is entanglement: two quantum objects, regardless of their distance, can remain linked. Measuring one allows us to know the state of the other, which can for example be the same or opposite.
The qubit constituting a quantum computer must be kept at very low temperature, a few steps above absolute zero, to function. However, heat is released as soon as information is extracted from the qubit. This challenge appears titanic when it comes to coordinating thousands, even millions, of qubits. No complete solution has been found yet.
Qubits are very sensitive, to the point where any interference causes calculation errors. It is estimated that quantum computers today make 1 error per 1000 operations. This rate would have to be divided by almost a billion times. We are currently trying to correct this problem by distributing the information over several qubits, repeating calculations or using classical computers to filter errors.
There is not just one type of qubit, but four chosen by different laboratories around the world and half a dozen being studied. We don’t know which one will win. The best-known model, used by IBM and Google, is called “transmon”, recently rivaled by fluxonium. The latter two are called “superconductors”. Certain experiments have also been carried out with the constituent element of current computers, silicon. Also used are the laser-excited “trapped ion qubit” and the “photonic qubit”. The latter, which requires a large number of lasers, has the advantage of operating at room temperature.
