The raw power of increasingly advanced quantum computers could necessitate advances in software to make sense of the noise.
Quantum computers certainly seem like strange devices. For humans used to living in a world driven by Newtonian physics, having a device dip into the world of quantum physics—where the rules are different and sometimes even counterintuitive—can seem inexplicable. And when those same devices actually solve complex problems and provide answers, it almost begins to border on magic.
Not too many years ago, there were still scientists who thought that quantum computing was a hoax. Quantum machines are built to run deep inside black boxes and must operate in a totally dark vacuum at temperatures close to absolute zero. So you can’t watch them as they work. They have to be designed that way, because their computing power is tied to putting atoms or electrons into a state called superposition, which is incredibly fragile. Almost anything can strip away that property and imprison atoms back into their normal, single state of being that makes up our Newtonian-physics based reality. Beams of light, heat, soundwaves, slight vibrations, air molecules or even radiation can devastate superposition in a process called decoherence.
These days, very few people doubt the existence of quantum computers. In 2019, Google, in partnership with NASA, achieved quantum supremacy by designing a quantum machine that could solve a problem that would have taken a traditional supercomputer thousands of years. That milestone puts the United States well ahead of other countries in the race to create powerful, and more useful, quantum computers.
In this country, most work on quantum computers is being undertaken by private companies and universities with heavy backing from the government. That is in contrast to most other rival nations like China and Russia, which are investing billions directly into government labs. Our approach seems to be working better. A recent report commissioned by the Department of Defense and conducted by the RAND Corporation shows that the United States leads the world in most key areas of quantum computing.
Most of the developments made so far in the quantum computing world have been because of improvements in hardware. Quantum computers use qubits, which are kind of like binary bits in traditional digital computers. They are powerful because a quantum device is designed to let the qubit—which can be something like a polarized photon or the spin of an electron—exist in multiple states at the same time. Instead of a digital computer’s bit that represents either a one or a zero, qubits can be both at the same time, plus everything in between. And having more qubits has so far equated to more computing power.
The Google quantum computer that achieved supremacy had 53 qubits. IBM recently announced a quantum computer with 127 qubits that is thought to be the largest in the world, although D-Wave is working on a new machine with thousands of qubits. There is some discrepancy about the numbers because of the vastly different ways companies can create qubits, but basically, more qubits means more power.
Software fixing hardware
However, while adding more qubits certainly gives more power, it does not make up for the inherent problems associated with quantum computers, with one of the biggest being that they are very prone to errors. Or, more accurately, they are difficult to understand and program so that errors don’t occur within their output. All quantum computers generate “noise” to some extent. They may return a correct answer to a question, but they will also send back a lot of useless junk, with the actual solution mixed in with it. Then it becomes a matter of trying to separate a needle from a haystack, or even a needle from a stack of other needles. Because of that, adding more qubits may not help the situation.
It's been suggested that artificial intelligence running on traditional computers could be employed to analyze the answers returned by quantum machines. That might make it easier to eliminate the noise more quickly than trying to do it by hand, but does not address the fundamental problem of inaccurate answers coming from quantum machines.
Instead of adding more qubits, the solution to this predicament might actually be software-based, letting programmers ask better questions so that noise is reduced or eliminated from the start. One of the reasons for all the errors is that the qubits can become entangled. This is a state where even if two qubits are physically separated, the actions of one can change the other. Albert Einstein amusingly described that property as “spooky action at a distance.” In practical terms, if you are accepting data generated from one qubit, but don’t know that it’s entangled with another, then there is a good chance that the data is being corrupted, but you may not know it.
Right now, scientists basically need to guess at how qubits are entangled and try to act accordingly. So it’s like trying to write a program to run on a machine where the rules are not completely known, and may change. Hence, a lot of noise gets returned with the results, regardless of the size of the quantum machine. And bigger machines could make the problem worse.
To try and compensate, scientists and researchers at The Massachusetts Institute of Technology recently unveiled a new programming language called Twist at the 2022 Symposium on Principles of Programming conference in Philadelphia. Right now, there is nothing quite like Twist. Most quantum computer programmers use assembly languages, or something like them, where they have to string a bunch of processes together without the benefit of much orchestration. They have to guess at the entanglements based on their observations of the data being generated.
Twist is designed to help scientists discover which qubits in their machines become entangled when working on a problem, and then take specific actions, like only accepting data from an unentangled qubit. The language of Twist mirrors other common programming languages and is designed to be easy for skilled coders to pick up.
“Our language Twist allows a developer to write safer quantum programs by explicitly stating when a qubit must not be entangled with another,” said MIT PhD Student Charles Yuan in MIT News. “Because understanding quantum programs requires understanding entanglement, we hope that Twist paves the way to languages that make the unique challenges of quantum computing more accessible to programmers.”
In the same MIT News article about the new language, Fred Chong, the Seymour Goodman Professor of Computer Science at the University of Chicago, talked about why Twist and other software developments may be just as important in the long run as putting more and more qubits into play.
“Quantum computers are error-prone and difficult to program. By introducing and reasoning about the purity of program code, Twist takes a big step towards making quantum programming easier by guaranteeing that the quantum bits in a pure piece of code cannot be altered by bits not in that code,” Chong explained.
As the hardware side of quantum computers continues to evolve, better software may be needed to help focus all of that raw power and potential. Twist may eventually seem like a small step towards that goal, but it’s undoubtedly a critically important one.
John Breeden II is an award-winning journalist and reviewer with over 20 years of experience covering technology. He is the CEO of the Tech Writers Bureau, a group that creates technological thought leadership content for organizations of all sizes. Twitter: @LabGuys