This article is more than 1 year old

Quantum computing: Confusion can mask a good story, but don't take anyone's word for it

It'll do great things in the future, just don't impose timeframes on when

Column It is said that the more you know about a subject, the less you trust newspapers about it. This is said because it is true.

It's not true for The Reg, of course, for The Reg, like any self-respecting specialist rag, employs specialist writers. Rather, it's when the national press – the sort with lifestyle and sport and TV reviews – cover anything technical, it often gets… icky. Journos call it failing the sniff test, when a story just feels wrong even before you go digging. You might not see immediately what the baby's nappy is full of, but by jove you've got a good idea.

So what to make of reports in the national press that Goldman Sachs, together with an outfit called QC Ware, is predicting that quantum computing (QC) will be commercially advantageous in financial markets by 2026? Is the first commercial use of QC a firm blip on the radar? What does it mean for us worker bees in our von Neumann-architected hives?

The story – based on a press release you can read here – has some intriguing details.

There is a big class of QC based on Monte Carlo simulation methods, which attack problems that have too many variables to exhaustively compute by classical methods. Monte Carlo changes variables randomly, which brings the whole answer space into view like a picture coming into focus, while classical computing can only bring it in pixel after pixel, line after line. The mathematics of Monte Carlo simulations say when you've got an answer you can trust, despite it being an approximation.

QC Monte Carlo simulation, though, is gruesomely complicated. Did you know, for example, that strongly frustration-free Hamiltonians can be designed to implement universal ground state quantum computation? Pay attention, there'll be a test next lesson. But QC Monte Carlo is a real and very exciting area of research – it, like many QC techniques, doesn't actually do anything useful because systems big enough to do good things are crippled by random noise.

Take it to the bank

Let's move onto complex financial derivatives. They make a few people lots of money, and occasionally cause global financial meltdowns because nobody understands them properly. Or rather, the risk you take on when you combine lots of different financial instruments is very hard to calculate. There are too many variables. But what's good about having lots of variables? A-ha! Bring on what the finance fiends call the "quantum advantage" – know more sooner, and win big with QC.

On the face of it, it is a fool's errand trying to sniff test a story about quantum computing and advanced financial tomfoolery when your sector knowledge of either is, shall we say, on a par with a housefly's knowledge of knitting. But there's more to a story than the facts.

It is possible to emulate machine learning and program your personal neural network to spot patterns in data where you don't need to know exactly what it represents. Google Images knows nothing about cats and kumquats. It couldn't tell you which one to stroke for pleasure or which to eat when you're hungry, much less why getting that wrong would be so socially awkward. But ask it for pictures of cats, and that's what you'll get. Ditto kumquats.

So let's have a look at the information surrounding the central claim. There are various good rules here. Does a press release use "could" a lot? Six times here: replace "could" with the functionally equivalent "might or might not", and it reads very differently.

Where does the story come from? Goldman Sachs loves selling financial derivatives, and the partner here, QC Ware, is a QC startup. Startups live on seed capital, and seed capital comes from convincing rich bastards that you can make them richer. And if QC Ware is talking about "in five years" – another warning sign, because you and they and the Count of Monte Carlo don't know what's happening in five years – so they'll need a lot of rich bastard geld to keep flying. But they're with Goldman Sachs, so must be pukka! And Goldman Sachs understands quantum, so it must be very clever too!

And quantum is a magic word that's shorthand for "science you're too thick to understand." Normally, "quantum" is an instant red card in a story, but here – well, there's a lot of real science behind QC Monte Carlo, so that gets a bye.

But the main thesis of the story – that even error-prone QC Monte Carlo could still manage a 10x improvement over today's systems in five years – ignores any improvement in classical tech over the same period. IBM, which also does lots of QC, has said that it can get to 2nm with silicon logic doing its deterministic thing, so let's see.

There's other good stuff in what's not said. Goldman Sachs bemoans how much it spends on slow derivative risk analysis, taking hours of top-notch computing to answer a question – and you thought Bitcoin was boiling the oceans? And just for once, it'd be good to see a press release predicting great things with quantum to have some bounds – error bars, percentage likelihoods, a bit of meat on the bluff. After all, they're claiming expertise in exactly this.

As quantum stories go, it's one of the better ones. But it's still hype, promising a chance of something in the mid-future that may not be that great even if it happens. That's news? No, it's self-promotion. If you want to get value from stories like this, dig in: follow the promises until you get to their sources.

QC Monte Carlo is genuinely interesting and will do great things, if QC ever does great things, but we don't know when. QC itself is genuinely interesting, even if it teeters on a sky-high pile of failed predictions. As the banker and the startup says – even if it's got lots of errors, it still might tell you something to your advantage. ®

More about


Send us news

Other stories you might like