I'm listening to the New Media panel at the MIT Enterprise Forum's Brave New Web conference in Boston with Laurie Baird (Turner Broadcasting), Jose Castillo (thinkjose.com), Alex Laats (PodZinger) and Andy Plesser (Beet.TV). I'm not that good at blogging live, but I've been noting a variety of interesting snippets.
I was particularly struck by Podzinger's media searching capabilities, not so much to index audio and video media, although that's useful. More interesting is the idea of detecting and avoiding spam in tags, thumbnails and so forth. One of the critical things Google does in evaluating the importance of web pages is evaluate the value of the incoming links. Today, content is being spammed by having every tag under the sun tied to one piece of content and by having deceptive thumbnails to lure people into (probably questionable) content. So it's possible the value of audio and video search technology will be in evaluating whether the tags associated with a media clip are plausible.
Alex Laats stated the average length of audio clips on the net is 22 minutes, but the average listening time is 3 minutes. Likewise, the average length of video clips is 7+ minutes while viewing times average 1 minute. I can think of all sorts of potential problems with these averages (presumably they include people who listen to just a few seconds and quit as well as people who listen to the whole thing). I need to catch up with Alex on where he got his data. Intuitively, there would appear to be value in being able to find just what you want within a media clip.
A common thread was that advertising will be the dominate model. Laurie Baird went further to project that, by 2010, 50% of the ad revenue would come from the 80% of less popular individual pieces of content. Along these lines, the best statement (I'm not sure who made it) was "copyright problems will be solved with new business models" — presumably advertising-based models.