I don't get the question? I know a lot of startups using all kinds of algorithms. Xobni(email analysis algorithms/bayesian analysis, statistical analysis), Topix (classification algorithm), Reddit (recommendation algorithm), automattic (spam detection/classification algorithm), krugle(ranking algorithm),joost(distributed system alogrithms) and the list goes on.
Most of the things published in 1st tier conferences are useless outside of academia. In my current field (computer graphics) this is glaringly obvious.
Trust me. Many of these algorithms are publishable in first-tier journals. There's a wealth of knowledge locked up in these proprietary Intellectual Property.
In terms of cleverness for dealing with spam, perhaps most founders are playing a game whereby they are hoping to be acquired before spam becomes unmanageable using simple techniques.
Because the vast majority of users won't appreciate them. Also users like predictability, particularly in social sites. Clever heuristics are not predictable.
Actually, I think pg was asking "on what do you base the theory that they don't use sophisticated algorithms?" rather than "why do YOU think that startups aren't using sophisticated algorithms?", but depending on the emphasis, it could go either way.
If they are using sophisticated algorithms, then they are using them behind the scenes. For example, such a startup might use sophisticated spam detection or a clever algorithm to get a performance boost.
This is different from a startup such as Google where PageRank is major breakthrough that is made obvious to everyone. We don't see many startups like Google in web 2.0. It seems that web 2.0 is more about harvesting people power and less about sophisticated algorithms.
I bet most people aren't thinking about eigenvectors when they are looking at Google's search results. The user does not (and should not be made to) care about sophisticated algorithms; only utility counts (in business, of course, not in research.)
Utility may count more (I would tend to agree with that), but at least some people will wonder why one product works 'so much better' than all the rest when it just came out of the woodworks.
"One can still ask the question though: of the startups acquired by major companies, how many of them use sophisticated algorithms?"
Yes, one could. Though it isn't the question you originally asked, and so you shouldn't be surprised when you don't get an answer to that question. The question you asked was "why aren't startups using sophisticated algorithms?" ;-)
Depends on what you mean by a major company. Does a startup acquired by Cisco qualify? Or Intel? Or Yahoo, even, in some cases? I know of a few small acquisitions Google has made where they purchased the research technology and not the founders!
