A 100% functional quantum computer is not yet available, although it is getting closer. However, the potential of computing based on this physics, capable of unraveling microbial dark matter (genetic material of microorganisms yet to be revealed), discovering new medicinal molecules, identifying each brick of a genome or optimizing a complex financial or industrial process, It is urgent to find shortcuts. BBVA researchers, which maintains a team specialized in this discipline with public and private participation, have achieved a distributed quantum simulation with classical servers and open source programming, replicable by any institution without the need for a supercomputer or a delicate computer based on the characteristics exotic from the subatomic world. That is, a way to do quantum computing with current technology, available and within reach of anyone.
The physical world we perceive is a trompe l’oeil, the shadows of Plato’s cave taken to the extreme. If we were able to jibarize ourselves to a subatomic size, we would perceive a dimension where we can be in two states at the same time (superposition), there is teleportation, energy is conducted without losses (superconductivity), there are frictionless flows (superfluidity) and a strange choreography marks the interaction of particles (topological order).
Unraveling this entire universe would allow us to answer such basic questions as what we are and where we come from, but also take advantage of its characteristics for practical applications such as quantum computing, with capabilities impossible to achieve by classical computing. The computer that allows the execution of quantum algorithms without errors still has a decade left, according to the most optimistic forecasts. Its main challenges are noise (a simple microwave or temperature alteration can ruin the process) and coherence time, the microseconds during which the superposition of states is maintained, which exponentially increases computing capacity.
However, there is a shortcut and it is the discovery made by BBVA researchers. “We have managed to simulate the execution of quantum algorithms using classical machines, scaling up to a total computing power of 38 qubits.” [bits cuánticos] and with the expected result in an ideal quantum computer,” summarizes Javier Recuenco, head of the Technical Architecture Innovation area at BBVA CIB.
“By doing simulation with classical computers, we have avoided the problem of coherence time and noise. I can now run the simulation for hours and hours,” he explains to add another fundamental element: “The algorithm grows with the number of qubits and I need more power. All of this has to be distributed in memory and we need a lot of it for this to work. Using a distributed quantum simulator becomes necessary.”
The new system does not aspire to surpass the capabilities of a completely fault-tolerant quantum computer, if this is a reality, but rather to take advantage of the advantages of quantum computing with the tools available now, despite the limitations. “It has a very high cost,” admits Recuenco in reference to the resources used for the proof of concept, the demonstration of the proposed method in the cloud, which on this occasion was from Amazon Web Service. They have stayed at 38 qubits, but they believe it is scalable.
A classical computer with 38 bits could only represent that many different states. However, the same number of qubits can simultaneously represent and manipulate 2³⁸ thanks to the property of superposition, which allows a qubit to be in a state 0, a state 1, or any combination of both at the same time. Therefore, a 38-qubit quantum computation can represent approximately 274 billion different states at the same time.
Distributed quantum simulation has a first application in portfolio optimization, risk calculation and finding the shortest path in graphs, a classic problem that seeks the optimal path between vertices or nodes. “But it can be applicable in any area. The universities must be very interested and the chemical or pharmaceutical industry. Or to find new battery components,” explains the researcher.
One of its great advantages is that it does not require a supercomputer or a network of quantum devices. According to Diego García Vaquero, director of architectures and co-researcher of the system, they started with devices with only eight gigabytes of RAM and reached a maximum of one terabit. A conventional network already existing in the cloud is enough. “And with open source,” he specifies. This premise is basic to facilitate the use of the developed simulation, which will be published in a detailed technical document so that it is replicable, as announced by the researchers.
Another advantage of the simulation achieved is that, by not depending on unstable systems, it can be executed in phases and establish what Recuenco calls “flags or intermediate control points” in the process to see how the algorithm progresses, as well as interweave the qubits without the topological limitations presented by real quantum computers.
200 times faster
This line of quantum simulation research is being developed by companies such as Fujitsu, which complements these developments with the world’s largest supercomputers and quantum computers. Last February, the company announced the development of a novel simulation technique, also based on distribution, that accelerates hybrid (quantum-classical) algorithms and reaches a calculation speed 200 times faster than previous simulations.
In the case of quantum circuit calculations using hybrid algorithms, larger scale problems require many qubits and days of processing. Simulations in the fields of materials and drug discovery can even require several hundred days.
Fujitsu’s technology enables large numbers of repeatedly executed quantum circuit calculations distributed across multiple groups to be processed simultaneously. Fujitsu has also devised a way to simplify large-scale problems with less loss of precision using one of the quantum simulators. In a single day, calculations are carried out that would take more than half a year with conventional methods.
Fujitsu believes that these models accelerate research on the practical application of quantum computers in various fields and are applicable to real quantum computers.
You can follow EL PAÍS Technology in Facebook and x or sign up here to receive our weekly newsletter.
