> Even more surprising, though, is the onset of glass behaviour in certain computer algorithms. For instance, if an algorithm is designed to seek out progressiv

> Even more surprising, though, is the onset of glass behaviour in certain computer algorithms. For instance, if an algorithm is designed to seek out progressively better solutions to a problem with a large number of variables it can become overwhelmed by complexity and grind to a halt before the optimal solution is found. By borrowing statistical methods designed for the fundamental study of glasses, however, such algorithms can be improved, and better solutions found. Are they talking specifically about simulated annealing here, or is this a more general observation?