Interesting article by Emil O. W. Kirkegaard, from 2023: IQ estimates of public intellectuals and personas.
The comments are arguably even better than the article and lend additional insight. My own take:
1. IQ estimates of historical figures are especially iffy. Newton is credited as co-discovering calculus, but what does this mean today in terms of IQ in a modern context? It is reasonable to assume he would score in ‘genius IQ’ territory if tested today with the benefits of modernity, but if we’re using achievements as a stand-in or proxy for IQ, then obviously it’s harder to say.
Conceptually, single-variable calculus is not that hard compared to something like string theory or algebraic geometry, so is Newton smarter than a string theorist? Can such a comparison even be made given they are products of different eras? String theory is harder, but string theorists have a much larger corpus of literature from which to construct the theory, whereas Newton had only his imagination and minimal literature to draw upon. This is even less reliable for non-science historical figures like philosophers.
2. Most of these estimates seem correct. As a baseline, eminent figures in the humanities can be expected to have IQs in the 125-135 range. This seems reasonable even without any biographic info or test scores, given that 4-year college grads typically can be expected to have IQs in the 110-120 range, and postgraduates or doctorates have somewhat higher IQs, maybe 125. The best people in their fields can be expected to be smarter than the median of said profession. An extra 5-10 points can be assigned to biologists, philosophers, or economists. Or an extra 10-20+ points for physicists, computer scientists, or mathematicians.
3. The apple does not fall far from the tree, but it does roll. Einstein and and Dyson were brilliant–their children, while also smart, much less so. There is significant regression to the mean going on even if IQ is highly heritable.
4. The estimates for the bloggers are also mostly correct. But the estimates for Curtis Yarvin and Eliezer Yudkowsky are too low. The latter, despite his fringe or nonsensical views about AI risk, was incontrovertibly a prodigy; his writings as a teen exist on the internet. He was obviously extremely articulate for his age, signifying way outlier verbal IQ and also outlier quantitative IQ. Same for Yarvin, another prodigy. We’re disregarding politics and only looking at signifiers of IQ. I put both their IQs at a minimum of 170, even if testing only maxes at 140. Scott Alexander is also much higher than his suggested score, too.
5. If we mean the highest possible IQ that can be reliably ascertained with psychometric testing or highly correlated proxies (e.g. the SATs or LSATs), then around 140 or so is the maximum, which can account for some of these scores being relatively unimpressive. Attempts have been made at creating higher ceiling IQ tests but you run into certain selection effects, in that the people who take these tests are not representative of the general population.
6. As someone in the comments notes, the estimates for many of these writers are too low, especially Noah Smith. In many ways, writing is harder than STEM, especially writing that goes viral and is consistently well-received. A lot of math is about rules that can be repeated, but success at writing is more unpredictable or subtle. Verbal ability is also as ‘g-loaded’ as math/quantitative. When people complain about calculus or other math concepts being hard despite doing well at more verbally-loaded tasks, it’s almost always a teaching issue or insufficient effort, not an intellectual one. I can bang out articles on a consistent basis, but the last criteria , not so much. Out of perhaps many hundreds or even thousands of pundits or commentators online, like on Twitter, presumably with IQs above 125 or so, to meet all those criteria of consistently publishing viral, well-received articles, is literally fewer than a dozen individuals, of which Noah Smith and Scott Alexander are among them.