That's what the good folks over at We Have Always Done It That Way are proposing: That meeting organizers shouldn't line up all their sessions six months before the conference, but just let the topics to be covered bubble up from the attendees on-site, and let the attendees be your speakers (a relatively new concept called the unconference). While I think this would work for some, mostly small, conferences, I can't see it happening for a major scientific convention, or other types of meetings where people do want to hear the latest breakthroughs from those who do the breaking-through.
But I agree that it's nearly impossible to determine what'll be hot in six months, in most cases. In the interest of compromise, here's the thought I posted over in their comments: Another possibility would be to put out a beta version of the program with topics you think will be of interest six months out, and let potential attendees, speakers, etc., comment and update the topics up until the day of the conference. That way, you can have a plan in place and still ensure that the content will be fresh and interesting, and speakers can target the information much more effectively. While I love the unconference thing, I don’t think many associations would be comfortable in taking that big a leap.
While, as David Gammel points out, this could backfire if your speakers aren't willing to listen and adapt accordingly, frankly, any speaker who wouldn't want to be as relevant and interesting to my specific group as possible wouldn't have a place on my program.
P.S. Anyone interested in making associations more relevant should check out We Have Always Done It That Way, and the book the authors are co-writing (and making available pre-publication in draft form to get just this type of feedback). Thanks to Ben for the pointer.