Of course, evolution is happening N-dimensionally, so going "downhill" in one aspect may be taking you "uphill" in another - which one wins depends upon selection pressure, luck etc.
And since selection pressures also happen in multiple dimensions (as well as being time and space varying - oh and non-linearities), the landscape of your objective function is constantly changing as well - so from one generation, "down" may become level or uphill, and genetic variation will also tend to scatter individuals through this N-dimensional hyperspace giving them slightly different starting points.
Enough for now?
"Today's robots are very primitive, capable of understanding only a few simple instructions such as 'go left', 'go right', and 'build car'." --John Sladek