Something I find really interesting (And dabbled a bit with at the university) is datamining. A very broad term that describes methods used to distill meaning from data. The Netflix prize is a good example of this: Netflix, an online video store wanted a good algorithm to determine what movies they should suggest to clients, based on their previous behaviour.
These suggestions have started to pervade more online circles. Facebook suggests friends and interests, Amazon suggests books, google autocomplete suggests certain search requests and even your mobile phone uses predictive input dictionaries. This is all an endeavour to make computers more intelligent. To allow them to learn from our past behaviour and extrapolate towards our future wishes. This is a good thing(tm).
I do however fear that these suggestions, because they are based on previous behaviour, will grow stale at some point. I believe that it is the nature of these things to grow more complex and become more and more adept at recognizing your behaviour and determining your classification. And maybe humans aren't as complex, rich and deep as we like to think, but I fear that being classified, I will get those things that suit me as I was, whereas i strongly advocate new experiences and contacts outside your personal field of view. An area which, in classification, is considered more error than intent.
As a humanist I wonder when we might iterate into the "What would be good* for this person" level of complexity, but that will require both a model of personal growth and a way to measure this. So for now, i would like to suggest ted talks, as i think all possible readers could benefit from learning and growth from that specific site.
* With good being : that which allows for the greatest personal growth. I said humanist didn't I?