Well, the AP can tell itself what to do, but they won't.
For the second part of your question, it's a chicken or the egg type of situation. Sure, the polling becomes better as more data points are fed into it, but how much of that is based on how the preseason polling set the positioning to begin with? We don't know what a "from scratch" November 1st poll would like in comparison to the "do we move this team up or down from their position in the previous poll" type we have now.
I also think that the polling "horserace" that is used now feeds too much into the committee's ranking process. They would never say that outloud, but it absolutely does.
I'm saying that college football would be better if we didn't rank teams until much later in the season, requiring whatever group is doing the ranking to examine a body of work, rather than preseason hype and conjecture, and then using that (obviously flawed conjecture) to base movement of teams through the rest of the year.