2 ms·
I don't really find this argument convincing. When considering a small number of circumstances, indeed the fitness function won't necessarily converge on truth
by naasking 9y ago
I don't really find this argument convincing. When considering a small number of circumstances, indeed the fitness function won't necessarily converge on truth because short cuts are probably more efficient in cognitive resources.
But once you start enlarging the possible set of circumstances, a lot of them will overlap in ways that seem random unless you discern the actual causal relationships between them. Without this process, then you're devoting resources to recording a never-ending list of correlations, but this list is strictly bounded if you instead infer causality.
In some cases, list of correlations will still be more efficient, but not in general when considering a large and complex world.