"Google's Book Search: A Disaster for Scholars" by Geoffrey Nunberg
As a cataloger, I find this article darkly amusing. It's an example of why good metadata can't (yet) be created quickly and automatically. Too many errors crop up, and bad metadata is as bad, or worse, than no metadata at all. If your metadata is bad, you just spent time and money creating something that is almost unusable the way it was intended to be used. As painful as it may be, it still takes time and the work of someone (often many someones) who knows what they are doing to produce really good, really worthwhile results.