Yeah, you can get news of the Google Video announcement from a horde of the usual suspects, but one photo by Cory Bergman crystallized why this is right in Google's wheelhouse. Keep in mind that I translate the Google philosophy of "organize the world's data", into "work on fscking hard data wrangling problems, at global scale."
There's a little mention by Bergman about breaking up videos into scenes. I don't know the state of video shot detection research, but I have to guess it's not ridiculously easy on "any old video". Hard problem 1.
Now that you've got the shots selected, and any other metadata you might have at hand, build up a good queryable index. Make it easy enough for television people to use. (I kid!) Hard problem 2.
Did I mention store all those videos and make them easily accessible anywhere on the net? While video has become more friendly with the Web, it still doesn't exactly play well. Partially because the files are huge and partially because of the bandwidth needed, both issues which are being eroded by Moore's law. But not at a pace to make dealing with video at a global scale easy. Ask Akamai. Hard problem 3.
I could see Larry and Sergey rounding up a bunch of those systems PhDs along with a few top web UI folks and saying, "We need to build a platform for the world's biggest video content management system. Every file's a couple of gigagbytes, they have practically no metadata, and we can't tolerate any latency delivering the media. What can you guys do?"
"And if it works, we all get rich again, brokering television advertising."
Yup, right up their alley.