None of these machines actually do anything useful for protein folding or even static protein structure prediction.
I have not seen any true examples of these computers generating approximate solutions for routing, scheduling or portfolio allocation that in any way justifies the cost of their R&D. I always read carefully when some company makes a deal with D-Wave about logistics optimization and it's always "this is something that could have been done on conventional computing for a reasonable price".
None the less, it seems like this photonic room-temperature version is a big step forwards vs D-wave enormous mainframes that required very low temperatures to operate. Practical usefulness is yet to be proven.
They might be able to solve those faster though (than silicon), since nobody has demonstrated a workable general-purpose (conventional) photonic computer - but TfA at least says they have a machine with 256 "spins", so while that's ~useless for real world problems, it seems to work while not being hugely complex, so there's hope for scaling it up, perhaps.
(I roughly agree with the other "useless" comments in this thread, with a caveat)
Note that Ising computers are quite special-purpose machines and can't (easily) solve general purpose computing problems.
The most known/hyped of them a couple of years ago was the D-Wave "quantum computer".
The machines can (approximately) solve useful NP-hard problems like routing, scheduling, resource/portfolio allocation, protein folding etc.
None of these machines actually do anything useful for protein folding or even static protein structure prediction.
I have not seen any true examples of these computers generating approximate solutions for routing, scheduling or portfolio allocation that in any way justifies the cost of their R&D. I always read carefully when some company makes a deal with D-Wave about logistics optimization and it's always "this is something that could have been done on conventional computing for a reasonable price".
None the less, it seems like this photonic room-temperature version is a big step forwards vs D-wave enormous mainframes that required very low temperatures to operate. Practical usefulness is yet to be proven.
They can't solve np hard problems better than conventional computers.
They might be able to solve those faster though (than silicon), since nobody has demonstrated a workable general-purpose (conventional) photonic computer - but TfA at least says they have a machine with 256 "spins", so while that's ~useless for real world problems, it seems to work while not being hugely complex, so there's hope for scaling it up, perhaps.
(I roughly agree with the other "useless" comments in this thread, with a caveat)