What does Google really want for YouTube?
by Josh Bernoff
The San Jose Mercury News, among others, is reporting that Google's YouTube will begin checking content for copyrights, specifically in a test with Disney and Time Warner's content. They finally admitted they are using Audible Magic's clever system for checking copyrighted music, which the Merc reported earlier but Google wouldn't confirm.
Now Google is asking copyright holders to send them copies of videos to check against. Mark Cuban, ever the copyright defender and YouTube attacker, cried foul at that one.
Now I'll tell you what's really happening here (in my analyst's opinion).
Google will indeed start "fingerprinting" and indexing any video copyright holders send it. It will even get around to using this to block video content that it can identify that way, just as it is promising. This blocking will be poor at first -- it's hard to recognize video content -- but it will get better over time. (It won't stop the Viacom suit, though.)
In the meantime, Google will start to assemble an (internal) collection of copyrighted video. No, they won't steal it or publish it. But as they develop algorithms for identifying video by machine, they can use this collection as a test. Does our algorithm correctly identify the last episode of "The Sopranos," the "Daily Show" from last Tuesday, the evening news from WMUR Manchester for April 1, 2007? Can we really tell when Cameron Diaz was the guest on the Tonight Show using our face recognition algorithm? And so on.
There are hundreds of thousands of hours of professional video produced every day all around the world. This adds up to way more than the 10 million or so music tracks Audible Magic is collecting -- by orders of magnitude. Audible Magic has a great asset -- all the music labels send it all their music for indexing. If Google gets its way, all those video producers will pipe their video into Google, with descriptions, every day. Every movie and all the DVD extras. Every TV show. Every commercial. Every baseball game and cricket match. Every music video. Everything. And that, my friends, is a pretty good repository to test your video search on.
This is just what Google needs to make a video search that doesn't suck. And since everybody's else's video search sucks, that will be an incredible edge for Google.