Just a bit of personal experience:
I looked into the Atkins/Paleo diet after Yudkowsky mentioned it, and the arguments were convincing enough for me to give it a try; I've switched to eating mainly vegetables and meat for the past couple months, with the occasional high-fibre bready thing thrown in (the fibre allegedly prevents the blood-sugar spike you get from carbs).
I am now back to my Summertime working-out waistline, despite pigging out on bacon-cheese rolls and being lazy as all hell. I feel healthy and fine - though I do miss carbs - and my muscle-mass is easier to generate, and has greater longevity.
I'm unable to locate the original post, but what Yudkowsky argued was that lack of rationality could lead to mortal effects, such as the politically-founded 'recommended diet' we've all been raised with; the history behind the food pyramid is more than a little suspect. I'd heartily recommend this diet for the rest of you (also, this diet should be a major consideration for present obesity levels).
Related to: Babies and Bunnies: A Caution About Evo-Psych, Superstimuli and the Collapse of Western Civilization.
The main proximate cause of increase in human weight over the last few decades is over-eating - other factors like decreased energy need due to less active lifestyle seem at best secondary if relevant at all. The big question is what misregulates homeostatic system controlling food intake towards higher calorie consumption?
The most common accepted answer is some sort of superstimulus theory - modern food is so tasty people find it irresistible. This seems backwards to me in its basic assumption - almost any "traditional" food seems to taste better than almost any "modern" food.
It is as easy to construct the opposite theory of tastiness set point - tastiness is some estimate of nutritional value of food - more nutritious food should taste better than less nutritious food. So according to the theory - if you eat very tasty food, your appetite thinks it's highly nutritious, and demands less of it; and if you eat bland tasteless food - your appetite underestimates its nutritious content and demands too much of it.
It's not even obvious that your appetite is "wrong" - if you need certain amount of nutritionally balanced food, and all you can gets is nutritionally balanced food with a lot of added sugar - the best thing is eating more and getting all the micro-nutrients needed in spite of excess calories. Maybe it is not confused at all, just doing its best in a bad situation, and prioritizes evolutionarily common threat of too little micronutrients over evolutionarily less common threat of excess calories.
As some extra evidence - it's a fact that poor people with narrower choice of food are more obese than rich people with wider choice of food. If everyone buys the tastier food they can afford, superstimulus theory says rich should be more obese, setpoint theory says poor should be more obese.
Is there any way in which setpoint theory is more wrong than superstimulus theory?