Ah, you graced us with another reply, while I was writing:
They are not improved by the random mutations per se. That seems to be the point you are missing. Mutations are of course, of their nature, quite random. But a mutation does not necessarily mean a step forward, an evolutionary progression. But, a certain mutation might enable the individual with the mutation to survive easier and thus reproduce more, perhaps passing on the mutation to the next generation. If it is something generally beneficial, it can catch on, and thus the species 'evolves,' a certain element about them changes. Now, it is a good question whether this principle could lead to such varied species as we have today, in the set categories we have, and why there aren't any in-between mutants as well, why are all the species so well-defined. If only a specific trait were beneficial, for example, why do you have several types of animals in an area, not just one? Of course, now it is in an equilibrium and there is often symbiosis between the different species, but why did they start diverging so in the beginning?