On the inadequacy of “human search” or “X is the new search”

It seems one of the new retarded valley startups brands itself as human-powered search. ma.gnol.ia or however you spell it claims/claimed that “found is the new search.”

The hubris and ignorance at work here is mind-boggling.

Suppose I am interested in how well a popular machine learning architecture has done on an important handwriting recognition dataset. I Google for “mnist svm“. Those letters are probably meaningless to you; mnist refers to a benchmark handwriting recognition data set based on data created by the National Institute of Standards and Technology; svm refers to support vector machines, an important, robust machine learning architecture that has recently been applied to a variety of computer vision tasks.

Do you know how fucking sparse the del.icio.us bookmarks are for things like this?

How are you going to pay/coerce a human to think up good results for important but obscure queries like this? Mahalo doesn’t even have any results for “support vector machine” when it’s one of the most important things to develop in the last 10 years in one of the hottest fields in computer science/electrical engineering/applied math - it just matches on “machine” and gives me shit about digital cameras and voting machines! This is worse than AltaVista in 1995! Jesus fucking christ!

If you aren’t a moron who queries for nothing but your own name and Apple products, you’re stuck with fancy information retrieval algorithms, at least until the Silicon Valley echo chamber thinks of something better. I’m sure when they do it will be simple to implement and scalable up to Internet scale. After all, when you’ve never written a line of code, even notoriously intractable problems like collaborative filtering are easily solved with a little bit of hand-waving and vague language.

“Oh,” they say, “but you have obscure interests. The vast majority of people searching just want to find pictures of Jessica Alba to masturbate to.” Yeah, but you add up all the obscure searches from people who actually use the Internet as a legitimate tool instead of just a TV substitute and you get quite a lot, some of it quite important — hey, isn’t that called the long tail or something?

Google handles most of these obscure queries pretty well, most of the time, as well as handling your searches for Dragon Ball Z episode torrents and poop videos. So, where the fuck is Mahalo or some other braindead thing adding value?