Readers of this bizarre story (in which a dubious claim about reflectivity of food in cooking transmuted into a flat-out wrong claim about the relevance of reflectivity of solar panels) might wonder how genius Nathan Myhrvold (Ph.D. in theoretical physics from Princeton at age 24, postdoc with Stephen Hawking for chrissake) could make such a basic mistake.
In an earlier comment, I dismissed this with a flip allusion to Wile E. Coyote. But now I’m thinking there’s something more going on.
In our blog discussion (see links above), Phil is surprised I didn’t take a stronger stance on the albedo issue after reading Pierrehumbert’s explanation. Phil asks: Why did I write “experts seem to think the albedo effect is a red herring” instead of something stronger such as, “as Pierrehumbert shows in detail, the albedo effect is a red herring”?
I didn’t do this because my physics credentials are no better than Myhrvold’s. And, given that Myhrvold got it wrong, I don’t completely trust myself to get it right!
I majored in physics in college and could’ve gone to grad school in physics–I actually almost did so, switching to statistics at the last minute. I could be a Ph.D. in physics too. But I’ve never had a great physical intuition. I could definitely get confused by a slick physics argument. And I suspect Myhrvold is the same way. Given what he’s written on albedo, I doubt his physics intuition is anywhere near as good as Phil’s. My guess is that Myhrvold, like me, got good grades and was able to solve physics problems but made a wise choice in leaving physics to do something else.
Now, it’s true, I don’t think I would’ve made Myhrvold’s particular mistake, because I would’ve checked–to start with, I would’ve asked my friends Phil and Upmanu before making any public claims about physics. In that sense, the difference between me and Myhrvold is not that I know more (or less) than he does, but that I have more of a clear sense of my areas of ignorance.
P.S. I’m on a Windows machine but my spell checker keeps flagging “Myhrvold.” I’m surprised that in all his years there, he didn’t use his influence to put his name in the dictionary. Then again, “Obama” gets flagged as a typo too. But “Clinton” it knows about. Hmm, lemme try some more: “Dukakis” gets flagged. But not “Reagan” or “Nixon” or “Roosevelt.” Or “Quayle.” If I were Nathan Myhrvold or Mike Dukakis, I’d be pretty annoyed at this point. Getting frozen out by Reagan or Roosevelt, fine. But Quayle??
Neither Nathan Myhrvold nor Mike Dukakis were ever president or vice president, so it's understandable that the office holders would be in the dictionary but the losing challengers or irrelevant to politics people would not.
Physics is hard, but sometimes it's also easy. For example Real Analysis is hard, there are all kinds of stupidly non-obvious technical conditions that might happen, but when you're dealing with physical models you can often say something like: if this technical condition in Real Analysis is important to the solution of my mathematical problem, then my mathematical problem is not a good model for the physical process, because the physical process simply can not have these bizarre properties. Famously Richard Hamming said:
"If whether an airplane will fly or not depends on some function that arose in the design being Lebesgue integrable but not Riemann integrable, then I would not fly in that plane" (https://secure.wikimedia.org/wikipedia/en/wiki/Richard_Hamming)
Often you can move along with your mathematical analysis, and if you find that in fact one of those special Real Analysis conditions holds, that's just a red flag to tell you to re-examine your physical model.
I imagine there are similar things in statistical analysis. If your model depends strongly on a normal distribution fitting out to the 99.999th percentile, then that's probably not a good model for whatever you're doing, or something like that.
Agnew also passes in (my version) of Word.
As does Gelman.
Oh, scratch that, I've probably added Gelman to my custom dictionary at some point. But I definitely haven't added Agnew.
I really want to comment more directly on this thread, but good taste suggests that instead I say that it is always an easy mistake to become over-reliant on models to the point of skipping essential experiments before committing yourself.
I think anyone might make a mistake. I would have hoped his physics training would have allowed him to recognize the mistake when it was pointed out to him (as in Pierrehumbert's letter you linked to last time).