Is there any sense in which this is solving NP-complete problems in polynomial time? They're clearly aiming at NP-complete problems, but it's not clear how efficient this machine is or how general it is.
They are developing system that could provide good heuristic solutions to NP-complete problem. That does not count as being able to _solve_ NP-complete problems. It won't find optimal solution in all cases, but it can be practically very useful.
This group at Stanford, collaborates closely with the Takesue group in Japan, who have the scaled version (2000 Node) ising machine in another article in Science:
5 comments
[ 2.7 ms ] story [ 15.7 ms ] threadhttp://science.sciencemag.org/content/early/2016/10/21/scien...
Their approach falls into the same category as simulated annealing or shrinking bubbles: https://www.technologyreview.com/s/512821/shrinking-blob-com...
http://science.sciencemag.org/content/early/2016/10/19/scien...