Many hard problems have huge search spaces, with nonlinear/multimodal cost surfaces, and little to no gradient information. Thus, single-solution methods often get trapped in local optima, lack robustness to noise, and fail to explore diverse regions. Population-based algorithms use a population of candidate solutions instead of searching with a single point.
These algorithms are often biologically inspired from concepts such as natural selection. the problem definition plays the role of the ecological niche, defining the fitness landscape.
