On smooth landscapes, any hill climber finds the optimum. On rugged landscapes with many local optima, diverse starting points and diverse heuristics dramatically outperform homogeneous teams. Different people get stuck at different local optima -- diversity of perspective enables groups to outperform individuals.
As landscape ruggedness (K) increases, the advantage of diverse teams grows dramatically. The more complex the problem, the more diversity matters -- because there are more local optima where homogeneous thinkers get trapped.
Each solver has a unique "perspective" (encoding of the problem space) and "heuristic" (search strategy). The best individual solvers may share similar perspectives and get stuck together. A random diverse group explores more of the solution space, finding higher peaks collectively even if individually less skilled.
Diversity of perspective + heuristics > raw ability. This holds when: (1) the problem is hard enough, (2) solvers are smart enough, and (3) the group is genuinely diverse. It is not that ability doesn't matter -- but diversity provides diminishing-returns-proof exploration.
With N ideas, there are N*(N-1)/2 possible pairwise combinations. As N grows, combinations grow quadratically, creating accelerating returns to knowledge. More ideas lead to more possible new ideas, which lead to even more ideas -- a positive feedback loop driving exponential innovation growth.
Innovation is combinatorial, not linear. The growth rate of new ideas is proportional to the number of existing ideas squared. This explains why innovation accelerates over time and why knowledge-rich environments (cities, universities) are innovation hotspots.