Common myths about quantum computers, and why they are dangerous
There are a lot of common misunderstandings about quantum computing, and left unchecked they could lead to some bad decisions
Quantum science and technology is a complex subject, and can seem magical and mysterious to many. This year, the International Year of Quantum, has raised awareness and interest in the potential that the technology holds for the future. This encouraged a thirst for more information, but unfortunately has also led a to a number of myths taking root. These myths are not only wrong because they allow those “in the know” to laugh at those who don’t know; they are dangerous, because they create misconceptions that, left unchecked, could drive bad decisions. The development of quantum technology will be a long term, global endeavour, but the right approach is needed from the start to make sure we maximise the benefits and minimise the risks.
So, in an attempt to set the record straight, in this article we’ll look at some of the biggest myths about quantum computers (future articles will look at some of the myths about other aspects of quantum) and discuss why they are wrong and why they are dangerous. While I won’t name names, it’s worth noting that all these have appeared in some form from sources that might be considered reputable - major tech vendors, big brand consulting firms, senior Australian government officials, and even official Government publications.
Myth 1: Quantum computers will run faster/better/cheaper than conventional computers
You’ll often see some sort of claim like this, which implies that eventually everything you can do on today’s computers can be replaced by doing it on a quantum computer. However, it’s actually likely that quantum computers will be vastly more expensive and slower, per operation executed, than the best supercomputers we have today. They will also be spectacularly bad at most things that supercomputers, or indeed any conventional (classical) computer does well.
Quantum computers will only be useful for a (probably narrow) set of computational problems for which we find conventional computers to be very slow or impractical to calculate. Even then, given the cost and slower clock speed of a quantum computer, for small versions of such problems it still won’t be worth using them. However, we believe some quantum algorithms will scale much better with problem size - so for a big enough problem and a big enough quantum computer there will ultimately be an “quantum advantage”1.
This myth is dangerous because it drives “magical thinking” - that you gain an advantage for any computational task by using a quantum computer. This leads to misunderstanding the potential benefits of quantum computers, and could even led to deprioritising investment in ongoing improvements to conventional computers.
Most importantly, this myth distracts from the need to work on finding the sorts of problems that quantum computers will be beneficial for, and designing the algorithms to use to solve them. Without focus on finding these use cases, all the effort that is being put into building a large quantum computer could be for nothing!
Myth 2: Quantum computers try all solutions at once
Even when people realise that quantum computers are best suited for solving particular problems, we see a lot of hand-waving about how they work. You’ve probably seen discussions on quantum probabilities and claims about how this means that the qubits in a quantum computer magically store all possible solutions in a superposition, and so can find the right solution rapidly, or even instantaneously. I won’t reproduce it here, but a common meme is multiple ball bearings in a maze game, exploring multiple possible paths through the maze at the same time.
This is wrong because it’s just not how a quantum computer works. Qubits do have more possible values than just the 0’s and 1’s that conventional “bits“ can store - you can think of the possible values of a qubit as points on the surface of a sphere - a bounded set but a large continuum of possible values. However, this doesn’t mean they can somehow store “all values at once”.
The possible values for a qubit actually correspond to quantum mechanical wavefunctions - conceptually you can think of them like physical waves such as ripples on a pond. With the right framing, sometimes you can set up these wave patterns can represent a particular real world problem, and design a quantum algorithm that makes these waves interfere with each other in a way that eventually creates a wave pattern that represents the solution.
The main danger of this myth is that it again drives “magical thinking” - implying that any problem that involves searching over many possible solutions can be solved with a quantum computer. One of the most pernicious examples is the common misunderstanding of the potential threats from a quantum computer to encryption. If you think the quantum computer can just guess all possible encryption keys at the same time, you end up concluding that all encryption is equally threatened by a quantum computer. In fact, the threat relates to some very specific maths used in some types of encryption, and hence the solution is some different maths.
Myth 3: Quantum computers process vast amounts of data at high speed
This sort of confusion generally comes from building on Myth 2 by assuming that convergence of technology buzzwords means that “quantum AI” must be a thing. The wishful thinking is along the lines that you can throw lots of data at a quantum computer, it calculates all solutions at once and so calculates the weights for an AI model of that data.
The truth is that actually quantum computers are very bad at processing vast amounts of data. Even the large scale quantum computers we envisage in the future might only effectively have 1000 logical qubits. Loading several terabytes, or even only gigabytes of training data will take a long time and be very expensive. There may be some specific types of problems where using a quantum computer for part of the AI model training and/or inference may give some advantage, but there’s certainly not going to be any general acceleration of AI from quantum computers processing vast amounts of data at high speed.
The danger of this myth is that it creates unrealistic expectations of the potential benefits of quantum computing. We might not know what the killer application is for quantum computers, but it’s unlikely to come from squeezing more out of the generic business cases for AI, some of which already rest on questionable assumptions. (I won’t try to get into the myths of AI here, as that would be a whole new series of articles!)
Overall, it is clear that that trying to give simple explanations of the complex field of quantum computing, without really understanding the details, can lead to serious misunderstandings and misconceptions. If these are not corrected, we will end up with an overly optimistic view of what quantum computing can do, causing major missteps in how we prepare to leverage the opportunities and mitigate the risks of developing such a transformational technology. It’s important, therefore, to raise awareness and make sure the development of quantum computing isn’t hit with a wave of disillusionment in a few years.
Even then, conventional computing will have a major role because large quantum computers will actually need large conventional computers to make them work.
most of the Myths are practical and true. however quantum computing has only advantages of solvng NP hard problems or algos requiring large computations e.g. space craft trajectory optimizations etc.
absolutely correct myths about quantum computing. most of teh classical complex problems ( space domain) may not be easily mapped on quantum circuits.
Another problem is prone to error.
Loss of coherence