Odd computer zips through knotty tasks.

Abstract

A century-old theoretical model of magnetism is giving rise to a hybrid computer, part classical and part quantum, that may capable of solving problems that overwhelm conventional computers. The so-called Ising machine, described in 100-bit and 2000-bit versions in two reports this week in Science , could tackle optimization problems that require finding the best solution among myriad possibilities, such as predicting how a protein will fold or allotting bandwidth in cellular communications networks. The machines take their name from the Ising model, which was developed in 1920 in an attempt to explain magnetism. Curiously, many optimization problems can be mapped onto the Ising model. Now, two overlapping groups at Stanford University in Palo Alto, California, and at NTT Basic Research Laboratories in Atsugi, Japan, have developed optical machines specifically designed to solve the model, at least approximately. Developers hope Ising machines may soon replace or aid conventional computers for some applications, although some researchers note it is not yet clear that the new machines can best ordinary computers.