Interesting idea. How effective it is depends. I still think that, especially when it comes to personal weblogs, a personal endorsement is going to be infinitely more effective than a canned ad. I think the only way this is going to be effective is to have ads that are carefully selected based on the content of your weblog. Even if automated, the ads need to be something that, if given the choice, you might actually endorse.
Actually, what would be ideal would be avoiding any automatic ad injection and instead buying placement. This, of course, requires more effort on the part of the author - but it helps the author's opinion come through a bit better, and still generates the same or better buzz for the product. After all, I subscribed to the feed to read the author's opinion, not a canned advertisement. I want the ad to reflect that. Of course, there should be some disclosure and the ad should be clearly marked as so. As Adam points out, Paschal does a lot of 'marketing' now (whether he gets paid for it, I don't know).
If we do go the route of automatic injection, though, I would want something like AdSense. AdSense uses the information that Google has about your site to deliver ads that are appropriate for the content of your site. Actually, I wouldn't be surprised to see this soon; they've already expanding into e-mail.
From a business model perspective, it could actually work. While many blogs have an impressive readership, the large majority of that is done via syndication. For example, of the people who read my weblog, less than 8-10% read directly on the web. This breakdown holds true both for posts to the aggregated site and posts only on my own weblog. Furthermore, I don't know how well a service like Bloglines does as far as reporting their own readership to .Text etc. (does it show up as a single reader?) - in which case the percentage reading via RSS would be even higher.