The future with quantum computers is almost here - but are we ready for it?

When approaching the time of the new, useful equipment, the human factor in the calculations becomes crucial

The future of computing is a huge metal tank? I will be disappointed if it turns out that inside it just sits a guy with a laptop and google.

Yorktown Heights, New York. I am in the room where one of the options for future computing is located. The computer itself does not make any special impression, and it looks like a metal tank hanging on the ceiling. The impression is made by noise - the periodic metal knocking prevailing in the room. This is the sound of the cooling system, bringing the equipment to a temperature close to absolute zero. And there is not an ordinary chip cooled there - this is IBM's approach to quantum computing.

In 2016, IBM made a sensational announcement by inviting the public to try out an early version of their quantum computer, which contained only five qubits — too little for any serious computing, but enough for people to gain real programming experience using a new technology. Technology developed rapidly, and IBM installed more tanks in its room with a quantum computer and added new readiness processes. The company has already expanded the project to 20 qubits, and optimistically announced that it is preparing a version with 50 qubits.


Like many of the company's early computers, the new IBM quantum computer is encased in a beige case.

When we recently visited the research center to them. Thomas Watson, the company's researchers, were more cautious about the statements, emphasizing that they made no promises, and that the 50-qubit computer was just another milestone on the way to the future of quantum computers. But they convinced us that IBM is guaranteed to participate in this future, in particular because of the peculiarities of the ecosystem that the company creates around its first attempts.

Chip building blocks

To create qubits, IBM uses superconducting wires connected to a resonator, mounted on a silicon substrate. Wires and the substrate allow the company to use all the experience gained in creating electronic circuits, but in this case the wire is a mixture of niobium with aluminum, which allows it to be superconducting at extremely low temperatures. Jerry Chau, who showed us the room for checking equipment, says that the company is still experimenting with details on improving qubits, and checking various formulas and geometry.

The resonator is sensitive to the microwave frequency, which allows you to assign or read the qubit value using microwave pulses. In each chip there are optical elements that receive microwaves as inputs and direct them to individual qubits. Microwaves themselves do not stand out by anything, so data entry is organized using existing, ready-made components. The only difficulty is to transfer the data to the input chip, deeply immersed in a tank with liquid helium. Equipment designed for this should not only withstand extremely low temperatures, but also survive when heated back to room temperatures. Although, being cooled once, the equipment is able to work endlessly, without requiring replacement.


Part of the cooling system contained in tanks. Liquid helium lowers temperatures to near absolute zero.

Quantum calculations are based on the entanglement of qubits. Chau told us that in order to entangle two any of the qubits, one can rely on the fact that their resonance frequencies are slightly different. If you refer to each member of a pair of qubits using the resonant frequency of its partner, then they can be confused. Then sets of pairs can be tangled in higher order systems. The coherence of qubits is preserved for 100 µs, but a couple of qubits can be confused in 10 ns. Chau said that it takes a few microseconds to tangle the chip, which gives enough time to prepare the entire system and perform calculations.

So if everything is so simple, why do we still not have a 50 qubit chip?


External cooling control panel - but not a quantum computer

The problem is that qubits are extremely sensitive to external noise. This may be the noise outside the device (although a metal tank helps shield the chip). It could be internal noise — the cooling system, the microwave cables, the chip components themselves — all of which can interact with qubits. And any interaction is catastrophic for calculations.

This means that any change in the chip architecture, even the addition of a single qubit, can potentially change the frequency and type of errors when performing calculations. IBM is engaged in careful modeling, trying to limit such problems before chip manufacturing, but, in a certain sense, this is an empirical and step-by-step process: you need to use the chip and see what happens. “Adding new qubits will allow us to identify sources of noise and crosstalk,” said Chau.

Sarah Sheldon, one of the scientists working on microwave systems that control and read data from qubits, says the same thing. “We have good tools for describing individual components, but there are no suitable methods for describing devices entirely,” said Sheldon. “By increasing the system, we are faced with situations in which the management of one qubit can lead to errors elsewhere.” Later she added: “We are approaching the limit, after which, classically, these devices can no longer be simulated - and how then to say, do they work correctly?”


And all for this? One of the test chips of a quantum computer

Superiority or volume?

The idea of ​​a quantum computer is that it can perform calculations of a certain kind drastically faster than ordinary computers. By using a sufficient number of qubits, a quantum computer can cope with tasks that a traditional computer would take longer to solve than the Universe exists. The threshold, having passed that quantum computer can do this, has been dubbed "quantum superiority". Google, making an announcement of its developments in the field of quantum computing in March 2018, mentioned this concept.

Naturally, Google formulated its concept of quantum superiority in terms of the allowable number of errors (the number that we would not tolerate from a traditional computer). IBM, on the other hand, develops error correction qubits. Unfortunately, this system requires several additional qubits. Bob Sator, IBM's vice president of quantum computing, suggested that a quantum computer with several hundred corrective qubit errors would require thousands of qubits to work.



Recall that the company is still working on a system with 50 qubits, not dealing with error correction. In the near future, quantum supremacy does not threaten us.

Instead, IBM suggests starting thinking in terms of "quantum volume", a measure combining the number of qubits used for calculations with the number of errors. Quantum volume would allow intelligently compare computers from IBM and those machines that are described by Google. However, this measure will not let us know how useful each of these machines will be.

And, to a certain extent, the answer to this question is, “It depends.” In some cases, you can safely accept errors. Decomposition of a large number into prime factors, for example, can be instantly checked on a classic computer. In other cases, errors will make the result of calculations unreliable, and there will be no easy way to check it. Therefore, current computers are a bit strange. “We can do something whose classically behavior cannot be predicted, and whose work is not spared from mistakes,” says Jay Gambetta, manager of the group for studying quantum computing and information at IBM. "We do not know what can be calculated with their help."

He noted that many classic computer algorithms were first created, and only then their effectiveness was proved. In the case of quantum computers it is quite difficult to prove something.

You can simply refer to the statistics: run the algorithm several times (potentially eliminating some or all of the advantage in quantum acceleration) and take the most frequent answer. IBM solves this issue, in particular, through public involvement, inviting everyone to try its computers. If they come in handy for something in the current state, there is a chance that someone would guess.

SDK for QC

How to attract the public to work on the machine, which requires an infrastructure that provides cooling with liquid helium, and which cannot perform already existing software? Part of the answer to the question is located at one of the walls of the computer room, in the form of a more traditional POWER microprocessor-based server. The server accepts tasks sent by people who subscribe to checking for quantum hardware. Among them are found both large financial firms, and students studying computer science.

But IBM relies on another part of the answer: the high-level SDK, which it calls QISKit. As Sara Sheldon, control system designer, describes this system’s microwave pulses rely on a set of arbitrary waveform generators, mixers and amplifiers. But QISKit allows users to abstract from all these details. It allows them to assign the initial state of individual qubits and their connections, and then software — something like a quantum compiler — translates this into a set of light pulses required for the system to work as it should. "You will never encounter microwave impulses," promises Jay Gambetta.

Programming takes place in Python, allowing people to use existing skills.


The cooling system is closed and helium does not have to be replaced.

Facilitating access is a surefire way to encourage third-party participation, but the success of the enterprise has set IBM the task of managing the community. Gambetta stressed that QISKit and its compiler were created in order to get a working system. He discussed the question of how to make the system more modular, so that you can include the contributions of various users in it - in his opinion, the IBM team has already accumulated more contributions to the code than it can handle. He also mentioned that he would like to start building something like code libraries, noting that such things as the implementation of the fast Fourier transform demonstrated their usefulness in solving a huge number of problems.

While IBM encourages such efforts, Gambetta also hopes that people will naturally wake up with interest in this topic. From the moment of birth, quantum computations mostly belonged to the field of physics. The computer scientists saw no reason to do this, because the equipment simply could not provide them with any calculations. The situation is beginning to change, and the contribution of computer scientists may be critical for the development of this field, since, as Gambetta notes, "they think of tasks differently, not like we, physicists."

He is also optimistic about the use of IBM equipment in computer science courses. When quantum computing becomes a normal part of educating people, it will be easier for them to treat them as an instrument useful for a specific set of tasks. At this point, a quantum computer will become something like a GPU or other specialized equipment, in the sense that people just need to decide whether it is worth the possible acceleration of efforts to write special code.

In general, after the visit, I got the impression that quantum computation reached the transition point. Part of the tour was needed in order to look at the equipment, but it may become the least interesting part of this area. With the progress of the movement and the addition of new qubits, the process will be a slow improvement and empirical testing. And the equipment hidden inside a carefully insulated tank filled with liquid helium is not the most interesting thing to see.

The whole task now comes down to how you can get the most out of existing equipment, and how it can be used to guarantee our readiness to increase the capabilities of computers. And at this moment, the human factor - a set of experiences and community management - is becoming increasingly important.