A few days ago I touched on the topic of computer virus evolution. I need to expand on that a bit.
For the most part, computer viruses are undergoing directed evolution. That is, there are people who are actively examining viruses, taking what they believe are the best features of each, and combining them to form newer, more powerful viruses. (This is analogous to sexual reproduction, since there is crossover of "genetic" material from multiple parents.) The viruses that are in fact more powerful go on to become the basis of yet more powerful viruses, while the weaker variants do not "bear offspring" and die out.
Natural evolution, on the other hand, lacks this critical directing faculty. Changes and exchanges are essentially random, and therefore proceed at a much slower pace than with directed evolution. (This is simplified somewhat - animals do select mates based on desirable traits, and therefore direct their own evolution to some extent. However, one can argue that this selection ability exists because it was beneficial to the organism - in other words, it itself was selected by evolution.) However, natural evolution has the potential to cover a much larger range of possible outcomes. For instance, a trivial and utterly non-obvious change to an organism's genetic code may make it more fit than the most carefully designed changes. This ability to exploit the non-obvious makes natural evolution ultimately more powerful.
Computer viruses will ultimately make use of both mechanisms to evolve. Errors in coding and in the reproductive process will add at least some of the random element to the otherwise carefully designed hybrids of existing viruses. Also, inspired (and not so inspired) changes that are not reflected in the current viral genetic pool will add a mutation-like mechanism.
As I have said before, it will eventually be in the viruses' best interest to be benign or even helpful. Consider real world viruses, like smallpox, for instance. Smallpox was very deadly to the host. In general, this is not conducive to survival anyway - if you kill all of your hosts, you will die yourself. However, in the case of smallpox, the hosts were humans, and this turned out to be fatal for the virus, which has been eradicated because it was too blatantly deadly.
At the other extreme, mitochondria, which are the "power sources" in all cells, are believed to have originally been virus-like organisms that entered into a symbiotic relationship with the precursor to modern cells. The cells provided a safe place to live, and the mitochondria provided energy as a kind of payment.
One further observation: it is possible to look at anti-virus software as a carnivorous animal that lives by eating viruses. Although the consumption of a virus does not provide it with energy or nutrients in the classical sense, it does provide the anti-virus software with a reason for remaining on the computer, perhaps to consume more viruses on another day. Thus, viruses and anti-virus software share a predator-prey relationship, and the more prevalent viruses become, the more anti-virus predators will be spawned to consume them.
You can respond to my ranting here.
Necessity is the mother of ranting.