1. Blogs aren't peer reviewed. Anything that is going to be referenced later needs to be reviewed by other people.
2. It is easy to change the content of a blog after the initial publication. This causes issues for referencing, and removes the "story" of how things were developed which is very important. Hundreds of ideas might end up being wrong before the right idea is found. We shouldn't be able to "delete" the ideas which were wrong from history. This is a part of the reason that papers don't reference websites (with some exceptions e.g. links to software tools or result databases). This is even taught to first year undergrads who try to reference blogs in their lab reports.
3. The journal acts as an independent trusted third party, blogs wouldn't have this. How do I know whether I can trust a blog owner?
There are probably more reasons. All in all I'd like a more diverse group of distributors but I don't think self-publication could replace journals.