Record-Low Error Rate For Qubit Processor 66
An anonymous reader writes "Thanks to advances in experimental design, physicists at the National Institute of Standards and Technology have achieved a record-low probability of error in quantum information processing with a single quantum bit (qubit) — the first published error rate small enough to meet theoretical requirements for building viable quantum computers. 'One error per 10,000 logic operations is a commonly agreed upon target for a low enough error rate to use error correction protocols in a quantum computer,' said Kenton Brown, who led the project as a NIST postdoctoral researcher. 'It is generally accepted that if error rates are above that, you will introduce more errors in your correction operations than you are able to correct. We've been able to show that we have good enough control over our single-qubit operations that our probability of error is 1 per 50,000 logic operations.'"
Still a long way away (Score:5, Informative)
An important thing to recognize is that most of this experiment was done with a single qubit. Practical quantum computing will need to have this sort of error rate for thousands of qubits. The good news is that the methodology they used looks very promising. They used microwave beams rather than lasers to manipulate the ions. This has been I think suggested before but this may be the first successful use of that sort of thing. As TFA discusses, this drastically reduces the error rate as well as the rate of stray ions.
We are starting to move towards the point where quantum computers may be practical. But we're still a long way off. In the first few years of the last decade a few different groups successfully factored 15 as 3*5 using a quantum computer. (15 is the smallest number which is non-trivial to factor using a quantum computer since the fast factoring algorithm for quantum computers- Shore's algorithm- requires an odd composite number that is not a perfect power. It is easy to factor a perfect kth power a bit by looking instead at the kth root. And factoring an even number is easily reduced to factoring an odd number. So 15 is the smallest interesting case where the quantum aspects of the process matter.) Those systems used a classical NMR system http://en.wikipedia.org/wiki/Nuclear_magnetic_resonance_(NMR)_quantum_computing [wikipedia.org] which has since been seen as too limited. There are now a lot of different ideas of other approaches that will scale better but so far they haven't been that successful.
One important thing to realize is that quantum computers will not magically solve everything. They can do a few things quite quickly such as factor large numbers. But they can't for example solve NP complete problems to the best of our knowledge, and it is widely believed that NP complete problems cannot be solved in polynomial time with a quantum computer. That is, it is believed that BQP is a proper subset of NP. Unfortunately, right now we can't even show that that BQP is a subset of NP, let alone that it is a proper subset. Factoring big integers is useful mainly for a small number of number theorists and a large number of other people who want to break cryptography. There are a few other cryptographic systems that can also be broken more easily by a quantum computer but there's not that much else. However, that is changing and people getting a better and better understanding of what can be done with quantum computers. A lot of the work has involved clever stuff involving using quantum computers to quickly calculate stuff related to Fourier series. Moreover, once we get even the most marginally useful quantum computers there will be a lot more incentive to figure out what sorts of practical things can be done with them.
So the upshot is that these are still a long way off, but they are coming. The way it looked in the late 1990s or early 2000s it was reasonable to think that the technical difficulties would make them never practical. They still are a long way from being practical but right now it doesn't look like there are any fundamental physical barriers and it looks like in the long run the problems that do exist will be solved.
Re:Progress! (Score:2, Informative)
Most early computing errors were caused by memory (not RAM as early technologies weren't random access) The shift from mercury delay lines to magnetic cores saw a serveral-orders-of-magnitude drop in error rates, and the associated increase in the viability of general purpose computing.