Here’s a thought I had yesterday on the train. Since video is hard to search (though there are many projects: Webseek at Colombia University, Altavista video search, SingingFish, Virage, Compaq’s Speechbot, …), classification may become more important with video than with the text-web.
Google’s image search (and most video search technologies) search on the text surrounding the video/image.
So here’s another idea: if I can link within a videostream (discussed here earlier), I can then annotate and make that content available to Google. So maybe my classification idea sucks - searchability may come from discussing video, not from classifying video. Discussing is, after all, more natural than classifying.