To explain (at least qualitatively) the unconventional magnetic behavior of certain materials, the physicists have been led to formulate and to study simple mathematical models. The concepts and methods they developed in this process appear to apply to a number of important random combinatorial optimization problems, for which they have proposed remarkable formulas. Their discoveries point towards a new branch of probability theory. Finding rigorous arguments to support their conjectures is a formidable challenge and a long range program, some steps of which are described in the present paper.

60K40, 82D30

random structures, mean field models for spin glasses