If Facebook has to be up there with the big boys, this is what they should be doing, hire a platoon of PhDs to do R&D, this is why biggies like Google, Microsoft, IBM, Oracle, Sun are very successful in innovations because they do have a platoon/army of PhDs that do nothing all day, except researching (either original or finding solutions existing problems from the knowledge already available in the literatures).

Facebook might already have a few PhDs I suppose, but having one, 2 or 3 PhDs is not enough. It doesn't mean that you have a PhD then all of a sudden you know everything in computing, no, that's not the case. Just the same as any discipline. It doesn't mean that when someone become a qualified physician , it automatically means that he knows everything clinical.

One just have to check out Microsoft Research, where they made most of their research papers (in various computing topics) available in different literatures. This is a sign of a company (Microsoft, Google, et al), that targets to be there in the long run (not overnight but years to come) lead by innovation.

If Facebook talks like a grown up boy, then it should act like once. Stop being a follower by implementing of what others are doing, just show leadership in innovation and hire a bloody army of PhDs to do R&D and progress innovation. In 2004, MIT Technology Review named Microsoft Research Asia the hottest computer lab in the world as described here. Now, that's what you call innovations.

Finally, I know that RWW is a fan of Twitter, where things like real-time search is often promoted as the next generation of search. Twitter is not doing realtime search. This is why Google, Microsoft & others haven't got into it, since it is almost impossible technologically at this stage. A search engine needs to do high intensive crunching every time a new document appears on the internet and these docs appear every few milliseconds or so. This process is impossible at this time. What Twitter is fact doing is just straight forward search of the raw data that is indexed, which appeared realtime, but suffer massively in its precision & recall capability (ie, it throws up a whole lot of irrelevancy). Search is about precision/recall and speed and it is funny that some uninformed internet commentators have said that Twitter is upto something that Google & Microsoft are still behind. The army of PhDs at Google and Microsoft have already know the problems to doing realtime search and there is no doubt that they have been researching on how to overcome those, but to say that Twitter which is a bunch of HTML cut& paste coders have developed something that the army of PhDs at Google & Microsoft have missed is an insult to the intelligence of people who know how search engine works. Again, to say that Facebook has developed some features that researchers at Google & Microsoft have missed is also an insult.

The only way Twitter could beat Google would be by developing a radically different search technology (a disruptive innovation). However, Twitter's search technology seems to have much of it's foundations in the work done by Google (at least I can infer this from Falafulu's comments).

So, I guess we ca't expect a Twitter Search to make a breakthrough, at least not soon.

Where can I follow your comments ? any backtype ? friendfeed ?

The rest is paralysis by analysis.

Factor-based: This is the Pandora model. Review elements on a set of factors, and use a distance formula to see how close the elements are to each other. The big issues here are you hope you've isolated all the factors that matter, and it's damn expensive to do this (hiring a lot of experts to review songs). I wonder how much of the rating has been automated by Pandora.

Similarity Scores: User-User or Item-Item, the Amazon, Netflix model. The user-user approach really only works well is if the correlations are high, something that is very unlikely due to people's varied interests (big problem with Digg's implementation). Item-Item seems to work a bit better, and the logic seems to make intuitive sense.

None of these models really deal with how people's interests change at any given moment. When I talk about this with people, they always say "Oh, so you are looking at what people list as interests and matching them to other people." That would be great if people updated their interests as they changed moment to moment. Who does that? I bet your Activities and Interest on Facebook are from when you first opened your account, how often do people really update them? A great recommendation model will understand this!