I think that's what Yudkowsky is essentially claiming, if I got it right - that true map is the one that gives correct description of the territory - meaning, if we make a prediction using this map and then actually interact with the territory and the prediction turns out to be correct, then it's the supporting evidence for the map being true, and vice versa. I derive it from the links given here: http://news.ycombinator.com/item?id=4961488

Good catch. I spent a long time thinking about this. I don't know who downvoted you, but you were on to something.