Genetic algorithms and human 'perfection' (Score 1)

As anyone who has worked at all with genetic algorithms would know, one of the major issues that arises when dealing with them is the exploitation vs. exploration problem. In other words, you need to be able to balance how your population combines the use of its current strengths with the search for new strengths. Although the point has been (and is still being) argued, while evolution cannot literally act in leaps and bounds, it does effectively do so.

This comes down to the "What good is half an eye?" argument that is frequently proposed in these situations. Evolutionary advances of this nature require time for evolution to build up the necessary components for the completion of the architecture. In other words, evolution needs be lenient enough for a series of detrimental mutations to produce an advantageous one.

Back to the major point... as in a genetic algorithm, life must be sufficiently tolerant of genetic imperfections so as to allow those 'imperfections' a chance of becoming something more. A fitness increase is often preceeded by a large downturn in fitness over preceeding generations. In algorithmic terms, how do we know that humans are not now on a 'local minima' in the global fitness space, and what damage could we do if we do not proceed with sufficient caution?

I'm sure that a qualified geneticist could rip this argument to shreds, but oh well...