If a problem is so difficult that only one supercomputer in the world can solve it, is it necessary to build a second, identical supercomputer to verify that the first one is working correctly and providing the right answer?
Both hardware and software developers routinely run build-verification tests on new products to verify they work correctly. They typically don’t require building or acquiring the equivalent of a second supercomputer to complete, however.
In the world of quantum computing, where quantum systems are expected to be so much more powerful than conventional systems that it might take weeks to replicate a result from the quantum unit, there has been no other real option until now.
Researchers at the University of Vienna may have solved part of that problem by weaving mini-problems, the answers to which are already known by the quantum system’s operator, into a quantum computer’s workload. If the quantum and conventional computers come up with the same answer for the miniature “traps” in the programming, it’s a much safer bet that the quantum unit is providing answers that are accurate, not simply too complex to verify, according to Stefanie Barz, lead author of a paper in Nature Physics describing the technique.
The “traps” are indistinguishable from the normal tasks for the quantum computer, to minimize the chance that it might identify them as anomalous and “correct” the result to match its own (possibly erroneous) result, Barz added.
Quantum computers aren’t likely to be any more imaginative or less accurate than conventional computers – which also undergo testing to ensure their results are accurate, but don’t usually require as great a cost in equipment or resources to verify a result as to compute it.
The test protocol, which applies to any quantum-computing architecture, also includes a process to verify whether the quantum computer is actually using quantum techniques in its calculations.
It has been oddly difficult to verify quantum computers actually do use quantum processes. In June a team of USC researchers published a way to confirm a 128-qubit processor in a D-Wave computer was making calculations based on quantum mechanical principles rather than those of classical physics.
The result goes farther than just testing whether all the components in a new bit of hardware are working correctly.
Because the process is independent of the hardware architecture of the computer itself, it offers the first relatively simple way to test both the results and the process of a quantum computer – without using as many quantum resources for the test as for the initial result, Barz wrote.
Image: Shutterstock.com/ agsandrew