I'll bet that the first iteration of this will largely suck -- any startup is going to need time to refine its algorithms and build up appropriate data sources for something like this. But the high concept is great. It probably won't find much that an experienced web-searcher couldn't find on their own, but by acting as an intelligent agent and doing the scutwork, it has the potential to save a bunch of effort and surprise you with some gems. Once they get all the details right, this could be really useful.
(And once they have the data sources right, I'll bet that they explore new modes of delivery. The heck with PDF and paper -- if I buy one of these guides, I want to be able to reach it from my phone, so that I have the restaurant guide at my fingertips. I'd be very surprised if they don't figure that out pretty quickly...)