Contrary to your claims, I think the original title is accurate. Our algorithm indeed “solves” linear systems of equations, although the answer is output in the form of the amplitudes of a quantum state, not in the form of a list of numbers. This is similar to the way in which MCMC methods find stationary distributions: they don’t output a list of numbers corresponding to a probability distribution, but they output a sample. In both cases, people may find solutions of this form unsatisfactory answers, in which case, it is possible to use these samples to estimate scalar functions of the solution.

This is made totally clear in our paper, and anyone who understands that our algorithm runs in o(N) time should understand that we’re not outputting a list of numbers: how could N numbers be output in O(log(N)) time? (As you point out, our algorithm is only interesting if the input is also created in o(N) time.) But because we wanted to avoid a potentially misleading impression, we changed the title to be more vague, so that people will only know what exactly we do with linear systems of equations by reading further into the paper.

As for the popular summaries of the article, probably they should have mentioned that there are many technical restrictions on the algorithm (related not only to the output, as you mention, but also the encoding of the input), that classical linear systems solvers are getting pretty close to linear time (and so our algorithm is only interesting with large, implicitly specified inputs), and that our algorithm relies on quantum computers which don’t exist yet. Probably the scientific press should generally be less optimistic in its tone. But your pessimism goes too far in the other direction.

If you want a better model of press coverage, this article does a really good job:
http://www.nature.com/nphys/journal/v5/n12/abs/nphys1473.html

But yes, I totally agree with the point you make about misrepresentation.

To be honest, i am currently a Ph.D. in Electrical Engineering. I was inspired many such “flawed” articles in popular press and decided to pursue quantum computation when i was an undergrad. Being young and naive i thought i could come up with another exponenitally faster quantum alogrithm! In the end i came up with a much simpler result which was published at WorldComp 2009 (where you were one of the keynote speakers) and infact realized the potential limitations of QC to solve a simple image processing problem and decided to move to another field.

It is then I once again came across the “fancy” news articles about Nanoelectronics and I came to USA and spent(loaned) 35,000$ to do my MS at a Top 10 school and realized that i reached a dead-end. No one knows where the research in Nanotube,nanowire or Graphene transistors is going! Looks like the industry will be still using CMOS in the near future.After my MS in so called Nanoelectronics, i did not get even a single job offer!!

Having burnt my fingers, i moved to more traditional ECE fields like Computer Architecure, Parallel Programming and Computational Intelligence during my Ph.D. studies and doing reasonably well.

A lot of people underestimate the harm caused by such frivolous reporting of scientific research.It can seriously mislead the young and naive aspiring researchers who will be the Scientists of tomorrow.

It is very unfortunate that it has become very fashionable to use the buzzwords ‘quantum’ and ‘nano’ very indiscriminately even by some of the scientific community as a means to attract funding and attention!

I always thought Science was an objective pursuit but unfortunately human beings are creatures of emotions rather than logic!

rrtucci’s last message seems to be talking about the effects of noise on the system, saying that continuous/quantum problems are the worst in terms of noise.

Although these are both true, I think they are completely different, and it sounds like you both think you were talking about the same issue.

Besides quantum Bayesian networks and independently of them, I use my blog to advocate for the use of quantum computers to do MCMC (Markov Chain Monte Carlo)

Let me call discrete problems (continuous problems, respectively) those whose set of possible answers is the natural numbers (real numbers, resp.). I believe both quantum and classical computers can produce statistical errors in their answers for both discrete and continuous problems. (Shor’s algorithm, though it solves a discrete problem, has a finite probability of failing, but there is a high probability of success after only a small number of repetitions of the experiment.) In the case of classical computers, we don’t normally worry about statistical errors (unless it’s a Monte Carlo algorithm) because current classical computers make such errors only extremely rarely. (If you operate then in a friendly environment, that is. You must, of course, worry about statistical errors if the environment is unfriendly; for instance, if your classical computer is operating at a very high temperature, or if it’s in outer space and exposed to very high levels of cosmic radiation.) Compared with classical computers, quantum computers have to contend with an additional and hard to handle source of statistical errors, namely quantum mechanical noise. One worries less about statistical errors for discrete problems (because the discreteness tends to filter out some of the noise) than for continuous problems, and less for classical computers than for quantum ones. Hence, you are right in your intuition of singling out as the worse possible scenario for statistical errors, that of solving a continuous problem on a quantum computer.