By definition, there's nothing really wrong with viruses. They're just self-replicating, that's all. If the cash in your wallet was self-replicating, you probably wouldn't complain. Virus researcher Fred Cohen has even put out a $1,000 bounty for the first developer who can come up with a truly helpful virus. So far, he hasn't paid out, but theoretically, a good computer virus is possible.
"Helpful" worms, however, may prove that even a "good" virus is a bad idea.
Helpful worms like Welchia, Den_Zuko, Cheeze, Mellenium and CodeGreen were designed in the name of helping the user. Welchia's design was actually kind of clever, finding and eliminating the Blaster worm by seeking out the same vulnerabilities as the Blaster worm, and then, usually, applying a security patch to keep any other worms from working their way in. The Welchia worm was programmed to automatically remove itself at a set date.
Here's the problem though: The main thing that worms do is slow down your network by feeding a constant stream of data through it. Whatever else they might do, that's the main thing people hate about worms. A helpful worm slows down the network just as much as a harmful worm will. Additionally, helpful worms are known to reboot the computer without the user's consent, which can be a major problem if you're right in the middle of a project that you haven't saved recently.
Helpful viruses are an interesting idea in theory, but they still self-replicate without the user's consent, they still eat up RAM and other resources, they still slow the network down. As technology advances we may see a day when helpful viruses are able to actually improve a computer's performance without any adverse effects. For the time being, however, there is that old saying about where the road paved with good intentions leads to...