A refreshingly measured and hype-avoiding review article on quantum computing targeted at non-specialists and without any equations was posted to arXiv last week: Quantum computing at the quantum advantage threshold: a down-to-business review.
Some quotes and my additional comments:
"Quantum computers should be seen not as competitors to classical machines, but rather as a supplementary class of devices aimed to tackle a distinct class of problems."
Think of a future quantum computer as being like a GPU or ASIC rather than a faster version of a general purpose PC.
[on NISQ algorithms] "For example, the estimation of the energy of a relatively simple molecule Fe2S2 would require as many as 10^13 measurements. Assuming (optimistically) that each quantum circuit run takes 10 ns, the single iteration would require about 24 hours."
Quantum computers won't solve hard problems instantly; they will solve (slowly) problems intractable using conventional computers. They probably won't be useful for optimizing the routes a delivery driver takes each day.
"In some cases, hopes to achieve quantum advantage by reducing an optimization problem to QUBO appear unviable altogether because of an exponential overhead associated with such reduction."
Another issue with some of the quadratic unconstrained binary optimisation (QUBO) mappings proposed in the scientific literature is they end up converting problems that are efficiently-solvable on a classical computer into an NP Hard problem that is much harder to solve in order to map it onto a quantum circuit; in effect taking one step forward and many steps backwards.
No comments:
Post a Comment