Google’s quantum computer achieves stage victory

100 million times faster than a PC, the Google machine solved a special optimization task, but the breakthrough to a wonder computer is still a long way off

A tidy sum, $15 million, was put on the table by Google and NASA in May 2013 – for a computer whose possibilities were still in the dark at the time. But the promise of the Canadian manufacturer D-Wave system lured the Internet company as well as the space agency: It is a certain class of quantum computer, specialized in solving so-called hard optimization problems. These calculators are called "adiabatic quantum computers".

Google and NASA are particularly interested in optimization tasks. What is it about? Often there are many solutions for a practical task, but they are of different quality. An example is "Problem of the traveling salesman". He wants to visit several cities by car. There are many ways, but some are faster, or shorter, or they are at favorable gas stations, etc.? Which one to choose?

Even with prioritization, the math quickly gets out of hand. With four or five cities, it can still be solved relatively easily with paper and pencil. But the complexity of the problem grows rapidly with the number of cities: even with 15 cities, there are tens of millions of possible ways of. Add just one city, and the number of alternatives multiplies. Even supercomputers quickly reached their limits with this exploding complexity. Finding the best or even a very good solution in acceptable computation time becomes impossible.

Google's quantum computer achieves stage victory

Hard optimization problems play an essential role at NASA or Google. The space agency wants to z.B. develop the most efficient logistics for their future missions. The problem of optimally using limited resources such as time, water or energy is extremely complex. Many of Google’s services are based on artificial intelligence. When machines learn to make optimal decisions autonomously or to recognize objects such as cars or faces in images, hard optimization problems often play a role.

The computer of D-Wave-Systems is a solution engine for optimization tasks. It has a chip with about 1000 so-called qubits. This involves superconducting conductor loops. In each loop the current can fly left or right around. In the language of computers, this way the bit values can be "0" (z. B. current around left) or "1" (stream right side up) encode. 1000 qubits can represent a myriad of combinations of zeros and ones.

Additionally, a magnetic field can be applied to each individual qubit, which interacts with the magnetic field caused by the current and gives the qubit a certain energy. This creates in the chip a "Energy landscape", which resembles a mountain range whose peaks have the highest energy and the thalers the lowest energy. In simple terms, optimization problems can be represented as such mountain landscapes. The solution is then to find the lowest point.

And thanks to quantum physics, the D-Wave computer should be able to do this faster than any normal computer. A conventional computer must proceed like a hiker, namely tracing the landscape to Fub and testing on the Hohenmesser to see if one can get lower in the next valley.

But in quantum physics there is a so-called tunnel effect that allows particles to get to the other side of an energy barrier without crossing it. As if all the valleys were connected by tunnels, which would save the wanderer a lot of time. (It should be mentioned here that another type of quantum computer uses other quantum effects to solve other tasks at lightning speed, such as cracking today’s common locks).

Google's quantum computer wins stage victory

Support system for the installation of the D-Wave Vesuvius processor, which has to be cooled down to 20 millikelvin. Image: Nasa

However, it is only an amption that adiabatic quantum computers can use the tunnel effect for a particularly high computing speed. The researchers have to try it out. So far, tests of the D-Wave machine have not shown any significant speed advantage, which has led to strong doubts about this type of quantum computer.

Hartmut Neven, a physicist at Google and responsible for testing the D-Wave machine, and his team have now programmed a particularly rugged mountain landscape into the quantum chip, as the advantage of the tunnel effect should come into its own here. For comparison, they ran a simulation of the operation of an adiabatic quantum computer on a conventional PC. In fact, the quantum computer came to the solution 100 million times faster than the PC. This proves that it can actually turn the tunnel effect into a speed advantage.

Not yet a wonder machine

On a "Miracle machine", As quantum computers are often dubbed, the world must continue to wait. This has three reasons.

First, the joke of a quantum computer is not just that it computes faster. This could also be achieved with many PCs instead of just one. Above all, it should not take much longer for a significantly more complex variant of a problem than for the simpler one. Whether it calculates the problem of the traveling salesman for 10 or 200 cities, theoretically it makes no big difference for it.

However, exactly that cannot be said on the basis of the present tests yet. It is still unclear whether a coarser version of the D-Wave computer with two or three thousand qubits can exploit the tunneling effect just as efficiently.

Secondly, the speed advantage achieved now is not in terms of a PC per se, but only in terms of the simulation that ran on it. This simulates the operation of the quantum computer, but has to do without the tunnel effect and was therefore slower. The researchers were concerned with this 1:1 comparison. For the tested optimization problem there are other solutions, which were even faster on a PC than the D-Wave machine has shown now. So the result has no meaning for practice.

Third: Hartmut Neven admits in a post on the Google Research Blog that the hardware of the D-Wave machine is not yet complex enough to program practically relevant optimization questions into it. For example, the control over the energy of the qubits had to be improved and the coupling of whole groups of qubits to each other had to be made possible. Zudem brauchte es eine deutlich grobere Anzahl von Qubits.

Nevertheless, Wolfgang Lechner of the Institute of Quantum Optics and Quantum Information in Innsbruck describes the result as "a very good one "promising". Physicist experiments with a similar system in Innsbruck. He is impressed that the speed advantage is already so large, even though the D-Wave machine has not yet fully developed "coherent" quantum computer, which means that it cannot maintain quantum effects over the whole computation. This is a hardware problem that future generations of quantum computers will overcome.

These did not necessarily have to use the same hardware as the D-Wave machine, he points out. The breakthrough does not have to come from California.