This is false. Natural selection doesn't
select at all.
Mutations occur randomly. Evolution doesn't cause them to come into being. Once a mutation has occurred, that mutation has a chance of being passed on to the next generation - this is genetic mixing in sexually reproductive species where the offspring only receives half of the material from each parent, so any given gene-level mutation has only a relatively small chance of being dropped. Consider that these initial mutations don't occur in every single sperm cell - they'll occur in one sperm cell out of millions. So *if* that particular sperm is the one that fertilizes the egg, then the initial mutation gets passed on, otherwise it doesn't.
Once past that initial event, the mutation has been passed to a single offspring, and is then present throughout their genetic code (cells divide and carry the same info with them). At that point, whether or not the mutation makes it to another generation depends on a lot of things, many of which are random - does the individual carrying that mutation get eaten before they can reproduce for example. It also depends on whether the individual can find a mate and have sex at all. It depends on whether or not the mutation is on the half of the genetic material that gets mixed in for the offspring. It depends on whether or not the fertilized ova develops into a viable infant. All in all, there's less than a 50% chance of the mutation being passed on to a second generation.
If a mutation actually confers a survival advantage, then the
probability* of it getting passed on is higher... but there's a massive amount of our genetic code that doesn't confer an advantage of any sort - but it doesn't **** things up either. By and large, most mutations get into the species not because they allow for an advantage, but because they don't reduce the likelihood to procreate.
Even more interesting, however, is when you look at how evolution functions at a population level. There are tons of simulations out there that have investigated the population effect - and we have evidence from the real world as well. Some detrimental mutations get passed on and become endemic in a population because the population is small and interbreeding is high, this concentrates the persistency of the mutation among the population even if it's disadvantageous. Similarly there are mutations that would be considered beneficial that do NOT become endemic in the species, because the breeding population is very large and by sheer chance the mutation gets lost or diluted to such a degree that it's immaterial.
It's also important to not underestimate the power of sexual selection as well. Peacock tails are definitely NOT a survival adaptation - they're a sexual adaptation for no other reason that that female peahens happen to like them. One might argue that intelligence in humans is an advantageous trait... but stupid people have more kids regardless of how advantageous we might think smartness is
Evolution has no mind, it does not select anything at all. The environment is the "selector" in that it's the gigantic pachinko machine that the genetic "marble" bounces through.
*ETA: Remember that random doesn't mean an even chance, it just means that it's stochastic as opposed to deterministic. Something that has a 90% chance of occurring, and does occur, is still a random occurrence - the occurrence itself is probabilistic in nature.