Paid searches (or how Google keeps us from curing cancer)?

Josh Hadro had a great blog post yesterday entitled "How Much Would You Pay to Search for the Royal Wedding Guest List?" This sparked what was going to be a little comment, but turned into a small rant. I'm not even sure if I believe myself 100%, or agree with myself 100% might be more accurate, but I do know that I definitely feel this way and it makes sense on the surface.

First, isn't paying for search exactly what libraries used to offer patrons- you know, back before the ubiquity of home internet? People were willing to pay for it then, but few, and they dropped it pretty quickly as soon as a free competitor was out there. As it is, people typically refuse to go get something that’s behind a pay-wall unless it’s the only option or they’re already going to subscribe anyway.
Why would search be any different? Unless Google (or some other company, to be fair) has some type of proprietary control over a subject and all related information, free searches will never disappear. Someone will always do it a little cheaper, with a few more ads, etc.

Second, did they quantify the tangential information gleaned during a physical book search- such as finding out a fact that was only slightly related at the time but added to the searchers' overall knowledge with a potential to use later? THIS, I believe, is the benefit of a less direct search.
It's the accidental discovery that we are losing- and it will hurt in the long run. These discoveries change the course of research. They change conclusions we draw. This is why library co-location vs. bookstore grouping makes more sense to me too, from a “for the good of man” kind of way. Do you get that with a targeted search? This is missing, especially if you’re really good at searching. The better search engines get at brining you back what you really want to see, the less there will be this cross pollination of information. I am sad about this.

Unfortunately, this is almost impossible to quantify. When you could it would have to be retrospective and anecdotal, so the data wouldn’t allow you to draw any quality conclusions anyway.