It all looks pretty flaky to me at this point, but I figure some of you must have run into her stuff before, and I was hoping you could share.

It looks like a biology-inspired, predictive approach somewhat along the lines of Hawkins' HTMs, except that I've not seen her implementation details spelled out as thoroughly as Hawkins'.

Her analysis seems sound to me (in the sense that her proposed model quite closely matches how humans actually get through the day), except that she seems to elevate certain practical conclusions to a philosophical level that's not really warranted (IMO).

(Of course, I think there would likely be practical problems with AN-based systems being used in general applications -- humans tend to not like it when machines guess, especially if they guess wrong. We routinely prefer our tools to be stupid-but-predictable over smart-but-surprising.)