" Does anyone – even Jonah Lehrer or the neuroscientists he consults – think that the brain, perhaps in cohort with the gut, is making decisions about how much we should eat, on how long we stay hungry and when we get full, so that we don’t overshoot by 20 calories a day. That’s matching intake to expenditure with an accuracy of better than 1 percent. (We consume, on average, about 2700 calories a day, so matching energy in to energy out and not overshooting by 20 calories requires better than one percent accuracy.)"
"...we actually have to be perfect in our matching of intake to expenditure or we’re going to get inexorably fatter (or leaner, if we err on the side of going hungry), or at least we have to average perfection over decades."
His argument is that the body is not accurate to within 20 calories a day. He appears unaware of the fact that this kind of regulation, which I agree isn't possible on a day-to-day basis (the brain never says "don't eat that last 1/12 of a candy bar"), is unremarkable when amortized over time. That is where his scientific inexperience shows. "Averaging perfection over decades" happens all the time and is perfectly unremarkable. He is oblivious to this fact.
This is a fatal error. Regardless of how it fits into his larger argument, he gets this basic and fundamental point wrong. When an author can't get facts like these straight, I become suspicious of the rest of his arguments. And indeed, a random sampling shows they are fundamentally broken. I would not trust him to explain any kind of complex science.