Dean posted this in his status, and I thought I'd bring it in here since he's always reminding everyone that we do have a forum for discussing things.
I think the video is pretty interesting, and brings up some aspects of advancing computers that often aren't talked about, though it definitely makes it sound easier to replace certain things than I think it will be in practice. For example law (since that's my profession and therefore what I'm familiar with); in the video most of the stuff it talks about isn't typically performed by actual lawyers. Sure sorting through huge amounts of data is something computers are good at, but the other things it mentions in passing, such as drafting documents, are more complicated than it makes out. A legal document is all about taking a position and constructing an argument to support it, something that would require very robust AI to do. This is not to say that a computer program could never be developed to do it, I'm sure it will happen at some point, but it requires advances well beyond that required to write a new story.
My other quibble is broader: sure it's silly to say "we'll just do other jobs we can't imagine yet" when you're talking about computers getting to the point of being actual, conscious AI, because at that point there's nothing humans can do that the machines can't, but I still don't think humans will necessarily go obsolete in the same way that horses did. The entire purpose of the economy is to provide for humans, and as computers can do more and more of the work that humans used to do the economy will change. So I guess rather than saying we'll do jobs we can't imagine, I'd say the economy will take a form we can't imagine, because humans won't actually be central to its functioning so we'll need entirely different systems for distribution of wealth, but there will be some kind of system.