It makes me wonder: how true is it that the web is too large to curate by hand? It's remarkable how fast a human can determine that a site is spam, and with the exception of a few large user-created sites once one has determined that a few pages on a domain are spam you don't have to look farther.

Kedrosky uses the number (unsure from where) 234 million sites. At $1/site you could have a number of quite well paid people checking off whether a site is 'real' or 'spam' at a price that Google could easily afford --- if it would improve their bottom line.

Even if you went page by page, you could figure out a way to build an index of spam-free pages. But would this really provide sufficient economic benefit? I think the answer is no, otherwise it would already have been done. What can be done to change this?