Waking up this morning, I can't help but think about how the imminent acquisition of Last.fm by CBS is just the next in a series of deals that financially validate the online social recommendation concept. (StartupSquad has some of the best news coverage of the deal. Last.fm blog post and comments worth a read as well.) I am very excited about the rumored acquisition of StumbleUpon by eBay as well.
"Users who liked what you've cumulatively told me you like, also tend to like these other things." It's a beautiful concept - I mean that I've been struck by the beauty of this concept across a number of sites for weeks. Everyone knows that's a big part of Amazon.com but it's also what makes StumbleUpon what it is, too. My favorite lately has been Pandora.
Recently I've heard people say things like "I worked on my Pandora 'stations' for months and I'm finally getting a really solid stream of music that I really, truly like." That kind of learning by a web service, starting from a point I designate and refining the trajectory based on thumbs up and thumbs down on subsequent movements, strikes me as fundamentally beautiful - especially when it's music we're talking about. I far prefer Pandora's interface over Last.fm's, by the way.
The roll of cumulative recommendation versus other core systems of analysis at Last.fm or Pandora isn't completely clear - but there seem to be two defining traits to both these sites and StumbleUpon: The interface can be related to very simply (though more complex use is also an option.) I get access to the fruits of my labor very quickly.
I used to use Furl.net for my social bookmarking - I miss it terribly, in fact. Furl would look at my bookmarks and suggest not just other URLs, which were less interesting, but it would recommend other users with similar interests. I could look at each of these and decide whether or not to subscribe to their bookmarks by email or RSS. Back in the day I chose email; I still get those emails and the signal to noise ratio is stunning, it's like a stream of pure gold.
It looks like MeFeedia offers something similar to this for video feed recommendations. The fact that del.icio.us does not offer recommendations seems a huge lost opportunity to me, almost a crime of neglect against my data. You know that companies that collect loads of my data are going to mine it for their benefit - I want to be able to do the same thing, at least on the simple level of getting recommendations relative to other users.
This post isn't terribly coherent or carefully crafted as much as it is a series of thoughts on the subject, but no series of thoughts here would be complete without the following. Service providers, give me access to my own damn data. I do the work using your tools, you hold the resulting data, you monetize that data for as long as I'm happy with you, I benefit from the act of data creation and secondary impacts like better recommendations over time. Then I find someone I like better than you and I'm out of here. Do you get to keep my data? Not exclusively, no! Keep it in aggregate if you'd like - but for goodness sake, if you think that holding my data hostage and threatening me with data poverty if I leave you is a way to keep me from leaving your service - well that's just a totally dysfunctional way to maintain a relationship.
Now I'm angry, thinking and writing about user control over our own data. VERY few companies are hip enough to this, I don't think any of the above discussed companies are. Why should they be until their users insist on control over our own data?
None the less, the CBS acquisition of Last.fm is a big validation of the social recommendation concept. I'm very excited about it and though I've got some big concerns, I am interested to see what a giant media company will do with it.
This looks great - a nice looking video player for films downloaded from Google Video, 
Advertising? Wasn't discussed. The examples the Google scientists provided fell into the following four categories:
The 
