Mar 23, 2022·edited Mar 23, 2022Liked by Erik Hoel
This is a fascinating theory, and I think you may be onto something here. I, too, have wondered about why there are fewer geniuses now than in the past. There are a few other potential contributing factors to this that I didn’t see mentioned:
(1) The rise of “publish or perish”. The aristocratic geniuses of the past could work at a leisurely pace and focus only on their most important ideas. Today, most intellectuals are academics or journalists, and both professions require increasing quantities of work from practitioners to maintain employment. This means most intellectuals are forced to spend more time on less important ideas that can be developed over the course of weeks or months rather than years.
(2) Geniuses are out there, but their contributions aren’t as obvious because of specialization. The intellectual world has expanded dramatically. There’s more specialized information than ever. As such, the contributions of geniuses will not be as obvious to the general public, or even intellectuals in their own field who work in different subfields. This applies even to the arts, where (I am told) to appreciate many modern art movements you need to be well educated in an array of specific artists and traditions.
(3) Geniuses are out there, but increased intellectual competition makes each individual less influential. There are so many more intellectuals today than even 200 years ago, and the competition is fierce. Take philosophy. The greats of the past engaged with only a relatively small number of thinkers, most of whom were dead. So a small handful of individuals had all the low hanging intellectual fruit to themselves. A philosopher today must compete with dozens or hundreds of thinkers, all of whom are hunting for objections to her work and/or rushing ahead to beat her to some of that low hanging fruit. Perhaps there are hundreds of hidden great philosophers today who, if there was far less intellectual competition, would eventually pick all the low hanging fruit themselves and therefore be more obviously great.
(I haven’t read the comments on the first post so apologies if these have already been addressed there)
This is brilliant because you've taken the conversation to a much more interesting place! The above discourse unfortunately went down the route of -- was this a mono-causal argument, what constitutes aristocratic tutoring, were all these geniuses aristocratically tutored, do we have a comparison-set of untutored would-be geniuses, etc.
But the main thing that felt lacking in the setup of Erik's original post itself was a demand-side view (i.e. the evaluation and categorization of "genius" by audiences). Your points 2 and 3 bring these aspects to the fore.
In fact, an extension of the kind of demand-side thinking evident in points 2 and 3 brings up a potentially fundamental tension in any analysis of demographic trends related to geniuses (genii?) --- Genius depends on scarcity. The darker the shadow, the brighter the perceived luminosity of the light. Thus, as a corollary: the more we have Brilliant & Original Thinkers (objectively measured) in our world, the less we have Geniuses (culturally categorized).
Not low hanging fruit. I think you are right about the leisure part. It is left out of modern intellectual pursuits. However, competition implies that people are competing for the same thing, rather than having individual thought. Genius is two things 1. powerful intellectuals and 2. original thinkers. Powerful intellect is not in short supply but groundbreaking originality is. Could this be due to the lack of leisure? Geniuses emerged not in a time of low hanging fruit but in a time when the synthesizing of knowledge was being competed for rather than preeminence in an already established field.
If you're willing to accept "parent did a little tutoring" as an explanation for some of the figures I named (eg Newton), I think that makes your theory much weaker than I had originally thought. In particular, it suggests that a big fraction of the kids being homeschooled by their parents should count as aristocratically tutored - and that's millions of people! I think that makes a lot of the rest of your post - about how this is almost unattainable today, about how it's impossible without unacceptable social inequality - fall flat. If ordinary homeschooling by parents counts, your headline result should be "home school your kids"!
(also, a lot of kids who aren't home schooled get some tutoring by their parents on the side, so I think tens of millions of children probably qualify as "at least as home schooled as Darwin or Dickens")
Likewise, if you're going to count Oxbridge as tutoring, your conclusion shouldn't be "too bad there's no aristocratic tutoring anymore," it should be "Go to Oxbridge, the one place still capable of producing geniuses!" And then you should compare Oxbridge to some equally selective institution like the Ivies to see if this really has an effect.
Although you never said the decline in genius was due entirely to tutoring, you implied it was a pretty strong factor. I think if eg 75% of past geniuses weren't aristocratically tutored, then the decline of aristocratic tutoring can only account for at most 25% of the decline in geniuses, which was not the impression I got from your post. Adding in the existence of current aristocratically tutored people, like home schoolers and Oxbridgeans, and adjusting for the percent of geniuses that they are, would lower these numbers further.
There's a whole section on parents acting as aristocratic tutors in the original piece. We see all sorts of historical examples of parents who set out to make their children geniuses, often with success (John Stuart Mill, the Polgar sisters, etc). But that doesn't look like today's homeschooling at all. Mozart being tutored by his professional musician father from before he could walk just doesn't seem to me like much contemporary homeschooling in the US, which is often for religious purposes, wherein kids follow along the set curriculum of regular grade school and their parents aren't professional teachers. I don't think it looks like Darwin's education, although his is, as I admit in the piece, a debatable case of aristocratic tutoring - he certainly had tutors and governesses, but how much one should count them depends on a lot of historical details of how involved they were that I simply don't know, since he also definitely attended regular school.
As for your other points about the Oxbridge method of learning: When Newton attended Cambridge at 17, he had a tutor, and most of the learning was tutoring-based, not lecture-based. From what I can tell, this is common in early universities, but (again from what I can tell), they all become more and more lecture-based as time goes on. So I'm not sure how much still goes on there, some people have said they still have a tradition of an unusual amount of tutoring. So I agree, go to Oxbridge! That is, after all, where Hinton went, your example of a modern genius. Perhaps not a coincidence.
As for your last point, of coming up with some number that represents the effect of tutoring on the production of geniuses: I'll be honest and say I have no idea what that number is. I think it's definitely non-trivial. If you look at intellectual life of Europe carefully prior to 1800, you find mostly aristocrats, and it was indeed the standard style to have tutors and governesses at home or at their elite universities. As I say in the piece, everyone who went on the Grand Tour took a tutor with them. That's like, most aristocrats! So I kind of doubt the numbers are as low as 25%-tutored. But I'd love to explore that further; some have suggested scouring through biographies using key words, or even AI, and comparing it to lists of historical geniuses, in order to get a better sense. That may sound easy, but done correctly I think that's basically the amount of effort of a high-tier science paper; perhaps a high standard for an initial blog post! But certainly, I have plans to do something like that in the future. Also, even if the number of the decline were 25%, it's still quite important, as it's something actionable we can do, versus the nothing we can about the "ideas are getting harder to find" hypothesis.
I'm slightly more optimistic about modern home-schoolers than you are (my neighbors are home-schooling their kid in a way that looks more like aristocratic tutoring than like a religious education or set curriculum, but they come from the same intellectual milieu we do so that might be less common than I thought).
I agree that the exact proportion of geniuses who were tutored depends on how far we are willing to go in accepting parents, one-room-schoolhouses, etc, and that it could plausibly be a big part of the effect, and that AI scraping of genius biographies would be a valuable cause.
I studied maths at Cambridge so can say a little about how it works these days. The content is taught in lectures, much like at other universities. The more unusual thing is that for each lecture course, you're assigned a supervision partner (another student taking the course) and a supervisor (a PhD student or researcher who understands the course well). Every two weeks, the lecturer sets an "example sheet" of problems for you to solve. You write up your answers and hand them in to your supervisor, who marks your work. You then get a one hour meeting together with your supervisor and your supervision partner, where you talk about your solutions and can ask questions.
That's a lot more small group teaching than you get at most other UK universities, and can be very valuable, but I don't think it's really comparable to how things worked at Oxbridge a hundred years ago. If you take 5 courses each term, then you only get 2.5 hours of "tutoring" per week, and there's a limit to the extent to which you can learn at your own pace when things are broadly tied to the fixed course syllabuses.
I think other subjects and Oxford work in a fairly similar way.
As a current Oxford finalist undergraduate in law, the primary modality of instruction is still 1-1 to 1-3 tutorials, supplanted by 40 hours of reading and a weekly essay in the tutorial subject. I only bothered to attend a lecture in person for the first time a few months ago - they are far from the primary mode of instruction in the humanities here, if that provides any insight. I think the tutorial system is very valuable, although I can’t really compare it vis-a-vis instruction at other universities
Erik, I think it was Justin H. Smith in February who pointed to the vastly different outcome expected of education before say, 1945 (to take an arbitrary date and primarily in the Western tradition). He was pointing to this in the context of plagiarism, and the argument proposed was that plagiarism is a thoroughly modern concept. Plagiarism was inconceivable in an education where repetition of known opinion was a virtue. Before modernisation, the outcome of education was to recall to the best of one's ability what others had written and thought around a common curriculum. In comparison we want every student to respond individually to every topic they come across. Imagine if we had the plagiarism rule in carpentry!
The point being that the academic tutoring is nothing like home-schooling, not because the techniques were different but because the education outcomes diverge wildly. Now we want an individual inquiring mind, and then they wanted conformity to accepted knowledge and tradition passed on in master to student arrangements. It is probably not at all ironic that the stability of conformity assisted gifted individuals to demonstrate ingenuity and genius, whereas the shifting sands of today's education preclude this. But that's just me being romantic.
It appears that one source of disagreement that is likely to keep recurring is due to an implicit assumption of "all else equal". Yes, education was different - but not just education. The content was different, as was the pedagogy. The expectations and responsibilities of the students were different. The relationships with their instructors, their peers, their parents were all different.
SA's comparison with parent tutors then and now, imagines a 21st century nuclear family transported several centuries. This is absurd, and comes across supremely strawman-y.
Meanwihle, your wish for a scientific conclusion kind of requires a certain assumption of "all else equal", which, in turn, legitimizes the above comparison.
Perhaps something you could do to avoid this cycle is to paint a more accurate picture of what aristocratic tutoring looks like. By expanding the definition of (x), you get better control of both x and the other variables.
Homeschooling spans a pretty broad range these days, and includes several million kids a year. As a result of growth, I think the share that's religiously motivated has shrunk to somewhere between 50-75% of the total. I bet you would find a meaningful number of kids that fit the tutoring profile, probably bigger than the number that were getting tutored in 18th century Britain or whatever. If there's no genius cohort coming out of homeschooling, potentially that casts doubt on the overall theory.
Well, yes, but there'd have to be good evidence that a significant percentage of modern homeschooling has the hallmarks of actual aristocratic tutoring - e.g., expert-level subjects at a young age, teaching not being based on academic observables like test prep, not just assigning the same schoolbooks at the same pace as regular public school, being given hours of engagement every day by experienced teachers, etc. From my knowledge of homeschooling (as someone looking into it for their own children some day) that's extremely unusual. For example, Bryan Caplan looks very close to this with his kids (who have their own podcast btw), but that's because he's a professor of economics who spends a huge amount of teaching them, i.e., he's "aristocratically" tutoring them. Professional tutors keep telling me that they aren't getting hired for positions like this, they are getting hired for test prep. If there were a bunch of heavily tutored (in this manner) homeschoolers, they'd be getting hired for them instead.
I was exclusively homeschooled until Harvard, and I've been reasonably successful but am no genius. I think there are two key differences between modern homeschooling and the aristocratic tutoring model. One is the level of expertise a parent can have versus assorted expert tutors. The other is just a level of entitlement we're just not too comfortable with right now, especially among the kind of people who are willing to take a huge amount of time to homeschool. When I went to Harvard I was astonished to see how my wealthy classmates felt they deserved to be on top, and in many cases ended up so, whereas I've never developed the same entitlement, which can help overcome that little inner voice of self doubt. So we can't really compare modern homeschooling with aristocratic tutoring except among the wealthy, and the intersection of those two groups is vanishingly small.
This thread seems to assume the people who are homeschooling for religious reasons (e.g. because they think school sex-ed pushes an immoral position) will therefore not give a good education, whereas people who homeschool for any other reason probably will. Why are we making this assumption?
I think the issue is that if people homeschool for ideological reasons, it is likely they will just throw away the parts of the curriculum they object to (say, sex education and Evolution) and add parts they favor (say, religion instruction). That CAN make the curriculum worse and doesn't do much for the overall efficiency.
If a family homeschools because it believes it will produce children more intelligent or knowledgeable (genius maybe?), at least they are trying to do what Mr. Hoel is try concerned about.
Unless the thing we're hoping they'll become geniuses in is sex ed, then excluding that from the curriculum isn't going to to affect the efficiency of making them geniuses. At least in my entirely conventional schooling the link between sex ed, maths, physics and a host of other subjects wasn't apparent.
My observation is that the parents who homeschool because they want to control the curriculum are extremely committed to their children's education, and seem as likely to produce geniuses as people homeschooling with sex ed included.
You are missing the point. What I said is, just throwing away parts of the curriculum you objetct to won't make the proccess more eficcient. It does not matter if it is sex education or (capitalist-inspired) economics (for those on the Left), it can be marginally better (arguably, teaching more Mathematics or Geography instead of sex education) and it can be worse if you get rid of something useful or replace it with something useless/false/bad.
Yes, "ideological" (it does not matter the ideological favor, it is just that religious homeschooling parents surely outnumber hippie homeschooling parents) homeschooling parents are dedicated to their children's educations. However, it is totally possible that they are primarily concerned with protecting them from "ideological/moral" harm, not necessarily making them smarter (i.e. they will be fine with their children learning as juch as the average pubkic school student). Is it hard to understand why they might never achieve something they are not trying to do for starts as opposed to someone whose whole point is trying to teach more/better?
You buying a gun might prove you are dedicated to your family's safety, it might even make it safer, but it won't per si make your yard look better so no one should be surprised if loving parents who buy guns to protect their children don't necessarily provide them better yards to play at than loving parents who have decided to improve the yard. It doesn't mean buying a gun makes yards worse, it just reminding us that not all gun-owners are yard-nuts.
I was under the impression that, at the very least, non-religiously home-schooled kids have a reputation for starting university at 14 or so, which while not the same as revolutionising their field certainly a dramatic improvement over baseline.
I wonder how much the format and rigidity of universities kills the advantages gained from earlier tutoring, since there isn't an alternative option for certification in our society?
I've looked at homeschooling a bit, and I believe those are often outliers. Similarly, you can find people on the other end of the spectrum who are angry they didn't learn much while being homeschooled. I think homeschooling makes it easier to be a positive outlier than the relatively homogenizing school system, but it's not a silver bullet.
>…Mozart being tutored by his professional musician father from before he could walk just doesn't seem to me like much contemporary homeschooling in the US, …
Unfortunately I think anyone who did that in this day and age in the United States (keep their child home from school and teach them almost nothing but music) would be hauled off to jail in no time.
It depends on the state, but I'm pretty sure that almost all states in the US would be fine with this happening. Laws are pretty lax on what you are allowed to do with homeschooling.
I really don’t have a clue because I’ve never dug into it. I live in New York State and I did look into homeschooling when my son was younger and there was a fixed curriculum that needed to be taught. I don’t know how arduous it was or how much time out of the available school time it would’ve taken.
The curriculum is designed for a child with average intelligence, learning inefficiently in a classroom full of children.
If your child is intelligent above the average and you can tutor them 1:1, it will take you about 1 hour to learn what the kids at school learn in 1 day. Then you can add 2 or 3 hours of whatever you want to teach, and your child will still have more free time than their peers at school.
As an example, children in the first grade learn alphabet and numbers up to 20. My daughter could read, write, and count up to 200, when she was five. So if we decided to homeschool, even if we did nothing, at the end of the first grade she would still be ahead of the average kids.
If you take the textbooks they use at school and do them at home, you can be finished in a month or two. The advantage is that you don't have to jump from one subject to another every day.
The challenge of homeschooling is time. Most people I know that homeschool, have one stay-at-home parent.
I think the more parsimonious explanation — what should be the null hypothesis — for the disproportionate numbers of aristocrats among past "geniuses", is that only they had the means and free time to while their lives away sketching equations, inventing finch beaks and discovering sonatas. This isn't to say I reject the idea that education played a part, but freedom from daily grind seems to be the magic ingredient to me.
Presumably we could test this by checking whether periods of 'total war' had an unusual level of technological progress (necessity is the mother of all invention) or an unusually low level (freedom is the mother of all invention), since in a total war far fewer people get to while their time away.
The golden ages of philosophy in ancient Greece and China both flowered in periods of political instability. Socrates was a veteran of the Peloponnesian War, and Mencius was an itinerant political advisor in the Warring States period. Both of these thinkers were in part responding creatively to the turbulence and uncertainty of their times in their thought. (The same can be said of a range of other great thinkers, from Hesiod to Hobbes.) That's some rough evidence at least that war can spur intellectual productivity, perhaps by disrupting the institutional stability needed to maintain "normal science."
Fair point. However, is designing incrementally cheaper and faster weapons "genius"? Is more scientists working together on the same issue, and thus arriving at a faster result, "genius"? I predict that if the government sank as much energy and money into researching [Inset Promising But Hitherto Theoretical Technology Here] as they did in the Manhattan Project, it could match the Manhattan Project; war incentivises governments to put in that kind of investment.
So war *will* look productive, if you count new military technologies (broadly defined to also include things like encryption methods or even medical advances), in a way that tells us little about what forces make more patent clerks come up with unexpected, groundbreaking theoretical models of gravity in peace-time.
tl;dr: Wartime inventivity might tell us a lot about fertile conditions for technological progress, but I don't think it tells us much about genius. Insofar as "genius" is a meaningful thing to talk about, those are different things.
On reflection I think your original point (in the past geniuses were disproportionately aristocrats, because everyone else was too busy not starving) is correct. However I don't agree that war time technological progress can be dismissed as 'incremental'. The argument by which the jet engine is only incrementally better than the prop, and the eletronic computer only incrementally better than the the mechanical computer, would equally declare general relativity to be 'incrementally better Newtonian gravity'. Therefore I would challenge anyone who deduced a possible corrollary of your argument - that if we let more people while the time away we'd get more genuises.
In the case of chess, I'd recommend looking into the case of the Polgar Sisters.
Their father was Laslo Polgar. He was a Hungarian psychologist whose focus of study was human intelligence.
He studied geniuses throughout history and came to the conclusion that geniuses were made, not born. He wanted to put his ideas to the test, so he married a Ukrainian researcher with them intent of having children and raising them to be geniuses.
He felt that anyone could become a genius in any field of sport, science, or the arts if you started your education at the age of three, and began to specialize at the age of six.
The field he and his wife choose for their children was chess, simply because it was easy to quantify success.
They had three daughters: Judit, Susan, and Sophia. All three went on to become the some of the most successful female chess players in history. 2 become grandmasters and the other was an international master. Susan has been the most successful collegiate chess coach in history, and Judit is widely considered to be the greatest female chess players in history.
Their story was very similar to Earl Woods and his son Tiger, and Richard Williams and his daughters Venus and Serena.
They were all geniuses within their fields. All were taught in a very similar way. None of their parents were experts in chess, golf, or tennis.
You can expand this to success in spelling and geography bees as well, where homeschooled kids are routinely winning, at least at levels far above their representation in the population.
I used to coach and compete in academic debate at the highest levels for years. I've seen it over and over where kids of successful coaches would become successful themselves, because they had such early exposure to the activity, and had tutor in the subject at home.
I don't think there is any doubt that this is the best way to educate someone. No different than how a tailor made suit or dress is better than one which comes off an assembly line.
However, we are now at a point where everyone alive today, for several generations, have all gone to "school". We are reflexively associate school with learning. Moreover, there is a massive system in place with a vested interest in the idea of "school".
School probably is better than the alternative for the vast majority of families who couldn't hire or tutor their children themselves.
However, slowly, there are people rediscovering the power of tutoring, although it is usually under the guise of homeschooling today. It simply works better and I don't think there is any debate about the matter at this point.
It might take an Elon Musk or some other high profile person to start to crack the social signaling edifice of schools. Or, something will need to replace the signal that elite schools offer.
"I don't think there is any doubt that this is the best way to educate someone. No different than how a tailor made suit or dress is better than one which comes off an assembly line." that's a great line, totally agreed
Britain's Great Man of literature Samuel Johnson has his early education described on his Wikipedia page as "Johnson displayed signs of great intelligence as a child, and his .. education began at the age of three, and was provided by his mother, who had him memorise and recite passages from the Book of Common Prayer. When Samuel turned four, he was sent to a nearby school, and, at the age of six he was sent to a retired shoemaker to continue his education. A year later Johnson went to Lichfield Grammar School, where he excelled in Latin. He excelled at his studies and was promoted to the upper school at the age of nine. At the age of 16, Johnson stayed with his cousins, the Fords, at Pedmore, Worcestershire.There he became a close friend of Cornelius Ford, who employed his knowledge of the classics to tutor Johnson while he was not attending school. After spending six months with his cousins, Johnson returned to Lichfield, but Mr Hunter, the headmaster, "angered by the impertinence of this long absence", refused to allow Johnson to continue at the school. Unable to return to Lichfield Grammar School, Johnson enrolled at the King Edward VI grammar school at Stourbridge.[25] As the school was located near Pedmore, Johnson was able to spend more time with the Fords, and he began to write poems and verse translations. However, he spent only six months at Stourbridge before returning once again to his parents' home in Lichfield."
Long-winded I know, however a good demonstration of the diversity of educational opportunities available in a society geared to education, even for the lower middle class. From the C13 on, charitable donations of schools and colleges demonstrate the significance of education, and being charitable schools, they often came with scholarships for those of ability to learn but lacking the ability to pay. The standard of learning expected was very high.
That is also a really good demonstration of how "aristocratic tutoring" bled into the lives of the non-aristocratic. Johnson's father was a bookseller as well.
"(a) science and arts are “mineable” or “exhaustible” in exactly the same way since the decline seems similar in both"
This isn't exactly how I would phrase it, so there's a chance I'm simply misunderstanding, but my immediate thought is that this isn't necessarily surprising. When you have optimization processes around (i.e. humans), you can have weird things happen.
Imagine a world where humans think in some abstract sense that science and the arts are equally valuable. But, the underlying rules of the world make science 10,000x easier than music (in some sense). People notice this, and so they award 1/10,000 the prestige for achievements in science, to account for it being less impressive. If humans chase prestige (instead of achievements directly), then they might pursue science in exactly the way to completely counteract it's advantage in easiness, and you might expect equal amounts of achievements, despite the underlying ease difference. The point is, with the right optimization pressures, what matters is not science's underlying "mineability" but instead how we, the optimizers, value it.
This creates a "control system" where the decline should mirror each other. If either declined faster, it would be harder, and therefore garner more prestige, and therefore get more attention, which shores up its decline, leveling the two out, and leading to a similar decline in both.
In fact, I would expect that without some sort of control system, it's difficult to explain b (the lack of impact of the explosion of free information), with *any* idea, or combination of ideas, (I mean, does the decline of tutoring manage better here? Even if it only wants to explain part of the issue, it doesn't seem to be coincident with the rise of the internet, so what happened at that time?) With the prestige control system it's trivial: Achievements got easier to make, so they became less impressive, so people stopped caring as much.
I think you mostly have the right of it, and I don't even know if I think the "prestige control system" is an actual thing, or just a useful category of thing to think about. But I just noticed that these two objections don't match my worldview.
Something similar, but not identical, is definitely true for athletics.
Human beings don't run very fast, and we aren't very strong for our size. But we have excellent stamina and an (almost?) unrivaled ability to throw things. A marathon runner or javelin thrower is infinitely more impressive among animals than a sprinter or a weightlifter. But nobody cares, and nobody should care. An achievement is always judged by comparison with other humans.
So even if art was in some cosmic sense much easier or much harder than science, the dynamics of the fields would still be similar.
Tutoring is indeed one of the few applications of AI I'm potentially excited about - that it could really help with tutoring and, more broadly, homeschooling. I do think (current) AIs are fundamentally passive - they don't *make* you learn anything, so it can never be quite the social/intellectual presence of a real tutor, but they are obviously potentially far more affordable.
I'm a bit of a pessimist when it comes to the post-work paradise of AI. I just don't see a mechanism for it other that direct governmental redistribution of companies that are so rich they can support UBI of an entire nation, and that seems unlikely, and additionally even if such capital existed the paths toward redistribution seems extremely, almost impossibly, murky to me.
I think you have highlighted one of the major challenges with the current state of AI's.
There might be something interesting that comes through with the ChatGPT plugin ecosystem:
1. Hi Chat, I really like birds and I need to learn new things in maths. Can you help me with some weird and wonderful things I can learn about maths through birds.
2. Sure, here is an outline of a few concepts:
2.1 Simulataneous equations
2.2 Linear algebra
I have scheduled these lessons into your maths slots in your diary for the week. And included a link to some highly recommended videos on the topics in the diary.
Still passive and not perfect, but a bit more engaged in getting the learner to engage with the created content on a schedule.
Let's watch the space and see what comes out of the wonderful world of AI + ed-tech
Although not getting much attention in the replies, the Bloom's 2-Sigma effect is essential to the argument since it helps us understand the mechanism beyond anecdotal correlation. Understanding the margins and mediators of that effect would be more useful, in my mind, than the Oxbridge-Ivy comparison that was suggested suggested. Linking here for those interested in the original research: http://web.mit.edu/5.95/www/readings/bloom-two-sigma.pdf
2. The expert performance research and deliberate practice adds detail the mechanism at play. Having a tutor or coach that can direct you to what you need to focus on, inspire you to practice, and provide individual, specific feedback is key. http://tashfeen.pbworks.com/f/EricssonDeliberatePracticePR93.pdf
Pleased to see discussion of the Oxford/Cambridge tutorial system. Particularly given the earlier start made on University in the 19th Century (and prior), I think it goes a long way to explaining the persistence of technically challenging and "elite" arts. It's mostly gone — no longer in STEM, and in a much reduced fashion in the humanities.
My sense is that the kind of talent needed to produce geniuses can be provided by aristocratic tutoring (in its different forms). But that same talent pool may be necessary, but not sufficient, for the actual production of genius.
At that point it seems like you need a confluence of historical and even species-level factors: there are "mute, inglorious Miltons" in many country churchyards, simply because the moment for that kind of poetry had passed.
The positive news is that (in my humble, but well-informed opinion) we are actually entering a new period where we will expect genius to flourish. In areas ranging from quantum computing and formal mathematics to political philosophy and cognitive science, we're entering a new period of low hanging fruit. Come join us!
I think the argument in your original post came across as more mono-causal than you intended, which seems to account for most of the disagreement between you and Scott.
Yeah, I agree, I'll take responsibility for that. But there's always this balance when writing these sort of pieces between (a) writing something that reads well, and (b) throwing in a bunch of caveats and hedges. If you do (b) too much you make yourself immune to criticism, but at the cost of being boring and making the pieces more longwinded and dragging. Frankly, I find a lot of online writers pick (b) and their stuff becomes. . . not a pleasure to read. Here, I probably strayed more toward (a) than (b), and in hindsight could have adjusted that ratio, but I also didn't know that Scott Alexander was going to pick over it.
Another reason your theory seems plausible (but would need more historical anthropology type work to back up) is that it would seem that tutoring would be a obvious educational style in societies over the past few hundred thousand years which didn't move into agriculture/hierarchy/class and therefore move people into specialization away from kids. If your way of teaching is 'do things with adults', which I expect most human societies have been, (i.e. learn what every plant and animal in an environment is, how they relate, and how to use them from elders) then it makes sense that you would have tons of geniuses - but before printing presses and writing and such, we wouldn't necessarily know about them, and their genius might be like "astonishingly good at hunting whales". Then with agriculture/inequality/etc, what used to be for everyone becomes limited to the rich - and the tutors got more specialized? Maybe? I've been reading Graeber (Debt, Bullshit Jobs, and Dawn of Everything) as well as a book called "Hunt, Gather, Parent", which I found very insightful besides the title, and which might have some good thoughts for this idea.
First, the empirical evidence in favor of elite tutoring over humdrum tutoring is not there.
That is, studies attempting to pit experienced against novice tutors repeatedly find no difference in learning outcomes.
The human tutors I've asked the last few months all agree with this. E.g.: "Yeah, there are some guys at my firm whom I don't think are very good, but their students' scores go up the same as everyone else's."
This is closely related to the interaction-plateau hypothesis in intelligent tutoring systems (very smart feedback-giving AIs only produce marginally better outcomes than fairly-dumb ones, so long as they're enough for the student to recognize their error), which is a major motivating drive behind the approach I'm taking for my own intelligent-tutoring work.
But, this can't be generalized too far: A typical study in this genre splits students into two groups, gives them a grad student who is either an experienced or novice physics tutor, and then compares their test scores. None of them look at 12 year-olds who have spent years with either a professor as a tutor vs. an undergrad, and see how many of them have started their own podcasts.
Second, apparently government requirements for mandatory learning are not as strong as people think they are, at least at the high school level, at least in some states. 100% of what I know about this I learned a few days ago from talking to the founder of https://www.powderhouse.org/ , who (slight caricature) told me that he can get students to satisfy all the state requirements for high school graduation by building some Github projects and then writing about it.
Third, I really want to see a more precise definition of genius by which we've declined.
I'll give an example of someone modern who was often called genius.
My friend, Michael B. Cohen, has published 15 papers in computer science since 2018.
This would be pretty good for anyone. But one makes this incredible is that he died in 2017 at age 25. He has over 30 publications total. https://scottaaronson.blog/?p=3468
From the stories I heard at his memorial, I'm pretty sure he attended normal public schools for K-8, and a public magnet school (Montgomery Blair) for high school.
We love these write ups! Also, thank you for mentioning our work rethinking education at the CollinsInstitute.org
One compounding factor that makes this kind of theorizing difficult is that your data could be framed a different way: Genius is not created by private tutoring but instead stamped out by (only) going through the industrial schooling system. This could be solved if we had some other schooling model providing a large set of data points but we don’t. In history the vast majority of people were either industrially schooled, tutored, or industrially schooled and tutored on the side (with “tutored” often being a catch all term for a huge range of pedagogies that are delivered on a one to one basis). (Perhaps how common people who got both industrial schooling and tutoring are on your genius list means a better way to word this theory is, "people who's only interaction with an educator is through the factor educating process during their developmental stage have their chance at being a genius significantly retarded".)
Or at least we did not have any alternative models being tried historically. One thing that really excites us at the Collins Institute (outside of the unique pedagogy we developed) is the rise of radical unschooling (essentially just letting kids do whatever they want), not because we think that unschooling is the optimal educational pedagogy (we don’t) but because the data it is going to produce could change the way we think about producing geniuses. Essentially, radical unschooling will produce a true baseline in education—”what happens if we do not interfere with a child's education at all”. The early data here (which unfortunately does not control for socioeconomic data) seems to indicate that these kids are outcompeting students of the industrial school model—essentially meaning that the industrial school system is “worse than doing nothing”.
Also, as someone else mentioned, László Polgár is a person to look at in this space. The fact that he was able to attempt to create geniuses within a specific modality and do so three times out of three indicates a better educational system would not be 20% better but could be something like two hundred times better.
Note: To your question of whether the rich still practice tutoring. Yes, they do. When I was growing up in the 90s I had tutors come by once a week in a couple subjects to explore them at a level beyond what the school was teaching me. I know that tutoring was their full time job so there must have been at least a few other families paying for tutors in the Dallas area in the 90s.
It should be noted that standardized testing is the opposite of tutoring. It is quite literally a technology created to eliminate the need for one on one attention.
The point is that if a family adopts homeschooling mostly for ideological reasons it is likely it will just throw away the parts it objects to (say, Evolution and sex education), which CAN make the curriculum even worse and does not do much for the efficency of the proccess. Say whay you will, but it is not sex education (alone) which is holding American students back.
Now, if a family adopts homeschooling because they think it will produce more intelligent or more knowledgeable children, it is at least teying to do what Mr. Hotel is concerned about.
This is a fascinating theory, and I think you may be onto something here. I, too, have wondered about why there are fewer geniuses now than in the past. There are a few other potential contributing factors to this that I didn’t see mentioned:
(1) The rise of “publish or perish”. The aristocratic geniuses of the past could work at a leisurely pace and focus only on their most important ideas. Today, most intellectuals are academics or journalists, and both professions require increasing quantities of work from practitioners to maintain employment. This means most intellectuals are forced to spend more time on less important ideas that can be developed over the course of weeks or months rather than years.
(2) Geniuses are out there, but their contributions aren’t as obvious because of specialization. The intellectual world has expanded dramatically. There’s more specialized information than ever. As such, the contributions of geniuses will not be as obvious to the general public, or even intellectuals in their own field who work in different subfields. This applies even to the arts, where (I am told) to appreciate many modern art movements you need to be well educated in an array of specific artists and traditions.
(3) Geniuses are out there, but increased intellectual competition makes each individual less influential. There are so many more intellectuals today than even 200 years ago, and the competition is fierce. Take philosophy. The greats of the past engaged with only a relatively small number of thinkers, most of whom were dead. So a small handful of individuals had all the low hanging intellectual fruit to themselves. A philosopher today must compete with dozens or hundreds of thinkers, all of whom are hunting for objections to her work and/or rushing ahead to beat her to some of that low hanging fruit. Perhaps there are hundreds of hidden great philosophers today who, if there was far less intellectual competition, would eventually pick all the low hanging fruit themselves and therefore be more obviously great.
(I haven’t read the comments on the first post so apologies if these have already been addressed there)
This is brilliant because you've taken the conversation to a much more interesting place! The above discourse unfortunately went down the route of -- was this a mono-causal argument, what constitutes aristocratic tutoring, were all these geniuses aristocratically tutored, do we have a comparison-set of untutored would-be geniuses, etc.
But the main thing that felt lacking in the setup of Erik's original post itself was a demand-side view (i.e. the evaluation and categorization of "genius" by audiences). Your points 2 and 3 bring these aspects to the fore.
In fact, an extension of the kind of demand-side thinking evident in points 2 and 3 brings up a potentially fundamental tension in any analysis of demographic trends related to geniuses (genii?) --- Genius depends on scarcity. The darker the shadow, the brighter the perceived luminosity of the light. Thus, as a corollary: the more we have Brilliant & Original Thinkers (objectively measured) in our world, the less we have Geniuses (culturally categorized).
Regarding #2 and #3: remember, in China, if you're one in a million, there are a thousand people just like you.
I think there's plenty of genius out there, we just have fewer people who are celebrity geniuses the way Einstein and Tesla were.
Not low hanging fruit. I think you are right about the leisure part. It is left out of modern intellectual pursuits. However, competition implies that people are competing for the same thing, rather than having individual thought. Genius is two things 1. powerful intellectuals and 2. original thinkers. Powerful intellect is not in short supply but groundbreaking originality is. Could this be due to the lack of leisure? Geniuses emerged not in a time of low hanging fruit but in a time when the synthesizing of knowledge was being competed for rather than preeminence in an already established field.
If you're willing to accept "parent did a little tutoring" as an explanation for some of the figures I named (eg Newton), I think that makes your theory much weaker than I had originally thought. In particular, it suggests that a big fraction of the kids being homeschooled by their parents should count as aristocratically tutored - and that's millions of people! I think that makes a lot of the rest of your post - about how this is almost unattainable today, about how it's impossible without unacceptable social inequality - fall flat. If ordinary homeschooling by parents counts, your headline result should be "home school your kids"!
(also, a lot of kids who aren't home schooled get some tutoring by their parents on the side, so I think tens of millions of children probably qualify as "at least as home schooled as Darwin or Dickens")
Likewise, if you're going to count Oxbridge as tutoring, your conclusion shouldn't be "too bad there's no aristocratic tutoring anymore," it should be "Go to Oxbridge, the one place still capable of producing geniuses!" And then you should compare Oxbridge to some equally selective institution like the Ivies to see if this really has an effect.
Although you never said the decline in genius was due entirely to tutoring, you implied it was a pretty strong factor. I think if eg 75% of past geniuses weren't aristocratically tutored, then the decline of aristocratic tutoring can only account for at most 25% of the decline in geniuses, which was not the impression I got from your post. Adding in the existence of current aristocratically tutored people, like home schoolers and Oxbridgeans, and adjusting for the percent of geniuses that they are, would lower these numbers further.
There's a whole section on parents acting as aristocratic tutors in the original piece. We see all sorts of historical examples of parents who set out to make their children geniuses, often with success (John Stuart Mill, the Polgar sisters, etc). But that doesn't look like today's homeschooling at all. Mozart being tutored by his professional musician father from before he could walk just doesn't seem to me like much contemporary homeschooling in the US, which is often for religious purposes, wherein kids follow along the set curriculum of regular grade school and their parents aren't professional teachers. I don't think it looks like Darwin's education, although his is, as I admit in the piece, a debatable case of aristocratic tutoring - he certainly had tutors and governesses, but how much one should count them depends on a lot of historical details of how involved they were that I simply don't know, since he also definitely attended regular school.
As for your other points about the Oxbridge method of learning: When Newton attended Cambridge at 17, he had a tutor, and most of the learning was tutoring-based, not lecture-based. From what I can tell, this is common in early universities, but (again from what I can tell), they all become more and more lecture-based as time goes on. So I'm not sure how much still goes on there, some people have said they still have a tradition of an unusual amount of tutoring. So I agree, go to Oxbridge! That is, after all, where Hinton went, your example of a modern genius. Perhaps not a coincidence.
As for your last point, of coming up with some number that represents the effect of tutoring on the production of geniuses: I'll be honest and say I have no idea what that number is. I think it's definitely non-trivial. If you look at intellectual life of Europe carefully prior to 1800, you find mostly aristocrats, and it was indeed the standard style to have tutors and governesses at home or at their elite universities. As I say in the piece, everyone who went on the Grand Tour took a tutor with them. That's like, most aristocrats! So I kind of doubt the numbers are as low as 25%-tutored. But I'd love to explore that further; some have suggested scouring through biographies using key words, or even AI, and comparing it to lists of historical geniuses, in order to get a better sense. That may sound easy, but done correctly I think that's basically the amount of effort of a high-tier science paper; perhaps a high standard for an initial blog post! But certainly, I have plans to do something like that in the future. Also, even if the number of the decline were 25%, it's still quite important, as it's something actionable we can do, versus the nothing we can about the "ideas are getting harder to find" hypothesis.
All right, all of that makes sense.
I'm slightly more optimistic about modern home-schoolers than you are (my neighbors are home-schooling their kid in a way that looks more like aristocratic tutoring than like a religious education or set curriculum, but they come from the same intellectual milieu we do so that might be less common than I thought).
Some people in my post suggested Oxbridge still does something tutor-y (see https://astralcodexten.substack.com/p/contra-hoel-on-aristocratic-tutoring/comment/5667724 ) but I don't know enough to compare it to past practices.
I agree that the exact proportion of geniuses who were tutored depends on how far we are willing to go in accepting parents, one-room-schoolhouses, etc, and that it could plausibly be a big part of the effect, and that AI scraping of genius biographies would be a valuable cause.
I studied maths at Cambridge so can say a little about how it works these days. The content is taught in lectures, much like at other universities. The more unusual thing is that for each lecture course, you're assigned a supervision partner (another student taking the course) and a supervisor (a PhD student or researcher who understands the course well). Every two weeks, the lecturer sets an "example sheet" of problems for you to solve. You write up your answers and hand them in to your supervisor, who marks your work. You then get a one hour meeting together with your supervisor and your supervision partner, where you talk about your solutions and can ask questions.
That's a lot more small group teaching than you get at most other UK universities, and can be very valuable, but I don't think it's really comparable to how things worked at Oxbridge a hundred years ago. If you take 5 courses each term, then you only get 2.5 hours of "tutoring" per week, and there's a limit to the extent to which you can learn at your own pace when things are broadly tied to the fixed course syllabuses.
I think other subjects and Oxford work in a fairly similar way.
As a current Oxford finalist undergraduate in law, the primary modality of instruction is still 1-1 to 1-3 tutorials, supplanted by 40 hours of reading and a weekly essay in the tutorial subject. I only bothered to attend a lecture in person for the first time a few months ago - they are far from the primary mode of instruction in the humanities here, if that provides any insight. I think the tutorial system is very valuable, although I can’t really compare it vis-a-vis instruction at other universities
Erik, I think it was Justin H. Smith in February who pointed to the vastly different outcome expected of education before say, 1945 (to take an arbitrary date and primarily in the Western tradition). He was pointing to this in the context of plagiarism, and the argument proposed was that plagiarism is a thoroughly modern concept. Plagiarism was inconceivable in an education where repetition of known opinion was a virtue. Before modernisation, the outcome of education was to recall to the best of one's ability what others had written and thought around a common curriculum. In comparison we want every student to respond individually to every topic they come across. Imagine if we had the plagiarism rule in carpentry!
The point being that the academic tutoring is nothing like home-schooling, not because the techniques were different but because the education outcomes diverge wildly. Now we want an individual inquiring mind, and then they wanted conformity to accepted knowledge and tradition passed on in master to student arrangements. It is probably not at all ironic that the stability of conformity assisted gifted individuals to demonstrate ingenuity and genius, whereas the shifting sands of today's education preclude this. But that's just me being romantic.
It appears that one source of disagreement that is likely to keep recurring is due to an implicit assumption of "all else equal". Yes, education was different - but not just education. The content was different, as was the pedagogy. The expectations and responsibilities of the students were different. The relationships with their instructors, their peers, their parents were all different.
SA's comparison with parent tutors then and now, imagines a 21st century nuclear family transported several centuries. This is absurd, and comes across supremely strawman-y.
Meanwihle, your wish for a scientific conclusion kind of requires a certain assumption of "all else equal", which, in turn, legitimizes the above comparison.
Perhaps something you could do to avoid this cycle is to paint a more accurate picture of what aristocratic tutoring looks like. By expanding the definition of (x), you get better control of both x and the other variables.
Homeschooling spans a pretty broad range these days, and includes several million kids a year. As a result of growth, I think the share that's religiously motivated has shrunk to somewhere between 50-75% of the total. I bet you would find a meaningful number of kids that fit the tutoring profile, probably bigger than the number that were getting tutored in 18th century Britain or whatever. If there's no genius cohort coming out of homeschooling, potentially that casts doubt on the overall theory.
Well, yes, but there'd have to be good evidence that a significant percentage of modern homeschooling has the hallmarks of actual aristocratic tutoring - e.g., expert-level subjects at a young age, teaching not being based on academic observables like test prep, not just assigning the same schoolbooks at the same pace as regular public school, being given hours of engagement every day by experienced teachers, etc. From my knowledge of homeschooling (as someone looking into it for their own children some day) that's extremely unusual. For example, Bryan Caplan looks very close to this with his kids (who have their own podcast btw), but that's because he's a professor of economics who spends a huge amount of teaching them, i.e., he's "aristocratically" tutoring them. Professional tutors keep telling me that they aren't getting hired for positions like this, they are getting hired for test prep. If there were a bunch of heavily tutored (in this manner) homeschoolers, they'd be getting hired for them instead.
I was exclusively homeschooled until Harvard, and I've been reasonably successful but am no genius. I think there are two key differences between modern homeschooling and the aristocratic tutoring model. One is the level of expertise a parent can have versus assorted expert tutors. The other is just a level of entitlement we're just not too comfortable with right now, especially among the kind of people who are willing to take a huge amount of time to homeschool. When I went to Harvard I was astonished to see how my wealthy classmates felt they deserved to be on top, and in many cases ended up so, whereas I've never developed the same entitlement, which can help overcome that little inner voice of self doubt. So we can't really compare modern homeschooling with aristocratic tutoring except among the wealthy, and the intersection of those two groups is vanishingly small.
This thread seems to assume the people who are homeschooling for religious reasons (e.g. because they think school sex-ed pushes an immoral position) will therefore not give a good education, whereas people who homeschool for any other reason probably will. Why are we making this assumption?
I think the issue is that if people homeschool for ideological reasons, it is likely they will just throw away the parts of the curriculum they object to (say, sex education and Evolution) and add parts they favor (say, religion instruction). That CAN make the curriculum worse and doesn't do much for the overall efficiency.
If a family homeschools because it believes it will produce children more intelligent or knowledgeable (genius maybe?), at least they are trying to do what Mr. Hoel is try concerned about.
Unless the thing we're hoping they'll become geniuses in is sex ed, then excluding that from the curriculum isn't going to to affect the efficiency of making them geniuses. At least in my entirely conventional schooling the link between sex ed, maths, physics and a host of other subjects wasn't apparent.
My observation is that the parents who homeschool because they want to control the curriculum are extremely committed to their children's education, and seem as likely to produce geniuses as people homeschooling with sex ed included.
You are missing the point. What I said is, just throwing away parts of the curriculum you objetct to won't make the proccess more eficcient. It does not matter if it is sex education or (capitalist-inspired) economics (for those on the Left), it can be marginally better (arguably, teaching more Mathematics or Geography instead of sex education) and it can be worse if you get rid of something useful or replace it with something useless/false/bad.
Yes, "ideological" (it does not matter the ideological favor, it is just that religious homeschooling parents surely outnumber hippie homeschooling parents) homeschooling parents are dedicated to their children's educations. However, it is totally possible that they are primarily concerned with protecting them from "ideological/moral" harm, not necessarily making them smarter (i.e. they will be fine with their children learning as juch as the average pubkic school student). Is it hard to understand why they might never achieve something they are not trying to do for starts as opposed to someone whose whole point is trying to teach more/better?
You buying a gun might prove you are dedicated to your family's safety, it might even make it safer, but it won't per si make your yard look better so no one should be surprised if loving parents who buy guns to protect their children don't necessarily provide them better yards to play at than loving parents who have decided to improve the yard. It doesn't mean buying a gun makes yards worse, it just reminding us that not all gun-owners are yard-nuts.
I was under the impression that, at the very least, non-religiously home-schooled kids have a reputation for starting university at 14 or so, which while not the same as revolutionising their field certainly a dramatic improvement over baseline.
I wonder how much the format and rigidity of universities kills the advantages gained from earlier tutoring, since there isn't an alternative option for certification in our society?
I've looked at homeschooling a bit, and I believe those are often outliers. Similarly, you can find people on the other end of the spectrum who are angry they didn't learn much while being homeschooled. I think homeschooling makes it easier to be a positive outlier than the relatively homogenizing school system, but it's not a silver bullet.
>…Mozart being tutored by his professional musician father from before he could walk just doesn't seem to me like much contemporary homeschooling in the US, …
Unfortunately I think anyone who did that in this day and age in the United States (keep their child home from school and teach them almost nothing but music) would be hauled off to jail in no time.
It depends on the state, but I'm pretty sure that almost all states in the US would be fine with this happening. Laws are pretty lax on what you are allowed to do with homeschooling.
OK.
I really don’t have a clue because I’ve never dug into it. I live in New York State and I did look into homeschooling when my son was younger and there was a fixed curriculum that needed to be taught. I don’t know how arduous it was or how much time out of the available school time it would’ve taken.
The curriculum is designed for a child with average intelligence, learning inefficiently in a classroom full of children.
If your child is intelligent above the average and you can tutor them 1:1, it will take you about 1 hour to learn what the kids at school learn in 1 day. Then you can add 2 or 3 hours of whatever you want to teach, and your child will still have more free time than their peers at school.
As an example, children in the first grade learn alphabet and numbers up to 20. My daughter could read, write, and count up to 200, when she was five. So if we decided to homeschool, even if we did nothing, at the end of the first grade she would still be ahead of the average kids.
If you take the textbooks they use at school and do them at home, you can be finished in a month or two. The advantage is that you don't have to jump from one subject to another every day.
The challenge of homeschooling is time. Most people I know that homeschool, have one stay-at-home parent.
Thanks for this. it's too late for me but its good to know. Maybe I can do it for my grandkids if my son gets busy. heh..
If I recall Caplan's descriptions of his sons' homeschooling correctly, not all that much time.
Yes, New York is one of the stricter ones, but even then you could get away with it
I think the more parsimonious explanation — what should be the null hypothesis — for the disproportionate numbers of aristocrats among past "geniuses", is that only they had the means and free time to while their lives away sketching equations, inventing finch beaks and discovering sonatas. This isn't to say I reject the idea that education played a part, but freedom from daily grind seems to be the magic ingredient to me.
Presumably we could test this by checking whether periods of 'total war' had an unusual level of technological progress (necessity is the mother of all invention) or an unusually low level (freedom is the mother of all invention), since in a total war far fewer people get to while their time away.
The golden ages of philosophy in ancient Greece and China both flowered in periods of political instability. Socrates was a veteran of the Peloponnesian War, and Mencius was an itinerant political advisor in the Warring States period. Both of these thinkers were in part responding creatively to the turbulence and uncertainty of their times in their thought. (The same can be said of a range of other great thinkers, from Hesiod to Hobbes.) That's some rough evidence at least that war can spur intellectual productivity, perhaps by disrupting the institutional stability needed to maintain "normal science."
Fair point. However, is designing incrementally cheaper and faster weapons "genius"? Is more scientists working together on the same issue, and thus arriving at a faster result, "genius"? I predict that if the government sank as much energy and money into researching [Inset Promising But Hitherto Theoretical Technology Here] as they did in the Manhattan Project, it could match the Manhattan Project; war incentivises governments to put in that kind of investment.
So war *will* look productive, if you count new military technologies (broadly defined to also include things like encryption methods or even medical advances), in a way that tells us little about what forces make more patent clerks come up with unexpected, groundbreaking theoretical models of gravity in peace-time.
tl;dr: Wartime inventivity might tell us a lot about fertile conditions for technological progress, but I don't think it tells us much about genius. Insofar as "genius" is a meaningful thing to talk about, those are different things.
On reflection I think your original point (in the past geniuses were disproportionately aristocrats, because everyone else was too busy not starving) is correct. However I don't agree that war time technological progress can be dismissed as 'incremental'. The argument by which the jet engine is only incrementally better than the prop, and the eletronic computer only incrementally better than the the mechanical computer, would equally declare general relativity to be 'incrementally better Newtonian gravity'. Therefore I would challenge anyone who deduced a possible corrollary of your argument - that if we let more people while the time away we'd get more genuises.
In the case of chess, I'd recommend looking into the case of the Polgar Sisters.
Their father was Laslo Polgar. He was a Hungarian psychologist whose focus of study was human intelligence.
He studied geniuses throughout history and came to the conclusion that geniuses were made, not born. He wanted to put his ideas to the test, so he married a Ukrainian researcher with them intent of having children and raising them to be geniuses.
He felt that anyone could become a genius in any field of sport, science, or the arts if you started your education at the age of three, and began to specialize at the age of six.
The field he and his wife choose for their children was chess, simply because it was easy to quantify success.
They had three daughters: Judit, Susan, and Sophia. All three went on to become the some of the most successful female chess players in history. 2 become grandmasters and the other was an international master. Susan has been the most successful collegiate chess coach in history, and Judit is widely considered to be the greatest female chess players in history.
Their story was very similar to Earl Woods and his son Tiger, and Richard Williams and his daughters Venus and Serena.
They were all geniuses within their fields. All were taught in a very similar way. None of their parents were experts in chess, golf, or tennis.
You can expand this to success in spelling and geography bees as well, where homeschooled kids are routinely winning, at least at levels far above their representation in the population.
I used to coach and compete in academic debate at the highest levels for years. I've seen it over and over where kids of successful coaches would become successful themselves, because they had such early exposure to the activity, and had tutor in the subject at home.
I don't think there is any doubt that this is the best way to educate someone. No different than how a tailor made suit or dress is better than one which comes off an assembly line.
However, we are now at a point where everyone alive today, for several generations, have all gone to "school". We are reflexively associate school with learning. Moreover, there is a massive system in place with a vested interest in the idea of "school".
School probably is better than the alternative for the vast majority of families who couldn't hire or tutor their children themselves.
However, slowly, there are people rediscovering the power of tutoring, although it is usually under the guise of homeschooling today. It simply works better and I don't think there is any debate about the matter at this point.
It might take an Elon Musk or some other high profile person to start to crack the social signaling edifice of schools. Or, something will need to replace the signal that elite schools offer.
"I don't think there is any doubt that this is the best way to educate someone. No different than how a tailor made suit or dress is better than one which comes off an assembly line." that's a great line, totally agreed
Britain's Great Man of literature Samuel Johnson has his early education described on his Wikipedia page as "Johnson displayed signs of great intelligence as a child, and his .. education began at the age of three, and was provided by his mother, who had him memorise and recite passages from the Book of Common Prayer. When Samuel turned four, he was sent to a nearby school, and, at the age of six he was sent to a retired shoemaker to continue his education. A year later Johnson went to Lichfield Grammar School, where he excelled in Latin. He excelled at his studies and was promoted to the upper school at the age of nine. At the age of 16, Johnson stayed with his cousins, the Fords, at Pedmore, Worcestershire.There he became a close friend of Cornelius Ford, who employed his knowledge of the classics to tutor Johnson while he was not attending school. After spending six months with his cousins, Johnson returned to Lichfield, but Mr Hunter, the headmaster, "angered by the impertinence of this long absence", refused to allow Johnson to continue at the school. Unable to return to Lichfield Grammar School, Johnson enrolled at the King Edward VI grammar school at Stourbridge.[25] As the school was located near Pedmore, Johnson was able to spend more time with the Fords, and he began to write poems and verse translations. However, he spent only six months at Stourbridge before returning once again to his parents' home in Lichfield."
Long-winded I know, however a good demonstration of the diversity of educational opportunities available in a society geared to education, even for the lower middle class. From the C13 on, charitable donations of schools and colleges demonstrate the significance of education, and being charitable schools, they often came with scholarships for those of ability to learn but lacking the ability to pay. The standard of learning expected was very high.
That is also a really good demonstration of how "aristocratic tutoring" bled into the lives of the non-aristocratic. Johnson's father was a bookseller as well.
"(a) science and arts are “mineable” or “exhaustible” in exactly the same way since the decline seems similar in both"
This isn't exactly how I would phrase it, so there's a chance I'm simply misunderstanding, but my immediate thought is that this isn't necessarily surprising. When you have optimization processes around (i.e. humans), you can have weird things happen.
Imagine a world where humans think in some abstract sense that science and the arts are equally valuable. But, the underlying rules of the world make science 10,000x easier than music (in some sense). People notice this, and so they award 1/10,000 the prestige for achievements in science, to account for it being less impressive. If humans chase prestige (instead of achievements directly), then they might pursue science in exactly the way to completely counteract it's advantage in easiness, and you might expect equal amounts of achievements, despite the underlying ease difference. The point is, with the right optimization pressures, what matters is not science's underlying "mineability" but instead how we, the optimizers, value it.
This creates a "control system" where the decline should mirror each other. If either declined faster, it would be harder, and therefore garner more prestige, and therefore get more attention, which shores up its decline, leveling the two out, and leading to a similar decline in both.
In fact, I would expect that without some sort of control system, it's difficult to explain b (the lack of impact of the explosion of free information), with *any* idea, or combination of ideas, (I mean, does the decline of tutoring manage better here? Even if it only wants to explain part of the issue, it doesn't seem to be coincident with the rise of the internet, so what happened at that time?) With the prestige control system it's trivial: Achievements got easier to make, so they became less impressive, so people stopped caring as much.
I think you mostly have the right of it, and I don't even know if I think the "prestige control system" is an actual thing, or just a useful category of thing to think about. But I just noticed that these two objections don't match my worldview.
That's an interesting thought. I'm not sure it's correct, but it's interesting - almost an "efficient marketplace for ideas" sort of notion.
Something similar, but not identical, is definitely true for athletics.
Human beings don't run very fast, and we aren't very strong for our size. But we have excellent stamina and an (almost?) unrivaled ability to throw things. A marathon runner or javelin thrower is infinitely more impressive among animals than a sprinter or a weightlifter. But nobody cares, and nobody should care. An achievement is always judged by comparison with other humans.
So even if art was in some cosmic sense much easier or much harder than science, the dynamics of the fields would still be similar.
> nature trumps nature
I assume one of these should be "nurture", but I don't know which one!
Ack! Great catch Isaac, I must have skimmed right over that. Should be "nature trumps nurture"
Hi Erik,
What do you think will be the outcome of having AI powered tutors available to everyone?
Khanmigo and Duolingo Plus (as well as the standard ChatGPT)
I am fascinated with allowing for a self-direct study of the world, with experts that come in and guide.
1. Will the experts be AI's?
2. Or will the abundance of new free time allow members of society to do this work for a fraction of the cost, and still have a decent living?
Thanks for the excellent writing and thinking about this space.
Thanks Reghardt! Great questions.
Tutoring is indeed one of the few applications of AI I'm potentially excited about - that it could really help with tutoring and, more broadly, homeschooling. I do think (current) AIs are fundamentally passive - they don't *make* you learn anything, so it can never be quite the social/intellectual presence of a real tutor, but they are obviously potentially far more affordable.
I'm a bit of a pessimist when it comes to the post-work paradise of AI. I just don't see a mechanism for it other that direct governmental redistribution of companies that are so rich they can support UBI of an entire nation, and that seems unlikely, and additionally even if such capital existed the paths toward redistribution seems extremely, almost impossibly, murky to me.
Thank you for the thoughtful reply Erik!
I think you have highlighted one of the major challenges with the current state of AI's.
There might be something interesting that comes through with the ChatGPT plugin ecosystem:
1. Hi Chat, I really like birds and I need to learn new things in maths. Can you help me with some weird and wonderful things I can learn about maths through birds.
2. Sure, here is an outline of a few concepts:
2.1 Simulataneous equations
2.2 Linear algebra
I have scheduled these lessons into your maths slots in your diary for the week. And included a link to some highly recommended videos on the topics in the diary.
Still passive and not perfect, but a bit more engaged in getting the learner to engage with the created content on a schedule.
Let's watch the space and see what comes out of the wonderful world of AI + ed-tech
Although not getting much attention in the replies, the Bloom's 2-Sigma effect is essential to the argument since it helps us understand the mechanism beyond anecdotal correlation. Understanding the margins and mediators of that effect would be more useful, in my mind, than the Oxbridge-Ivy comparison that was suggested suggested. Linking here for those interested in the original research: http://web.mit.edu/5.95/www/readings/bloom-two-sigma.pdf
Two other pieces of research are also relevant:
1. Raj Chetty's works on Lost Einsteins suggests that have models of inventors is key to children growing up to be inventors. Being tutored by a Kelvin or a Descartes may have similarly inspired unconventional aspirations to do great things. https://mitsloan.mit.edu/ideas-made-to-matter/lost-einsteins-us-may-have-missed-out-millions-inventors
2. The expert performance research and deliberate practice adds detail the mechanism at play. Having a tutor or coach that can direct you to what you need to focus on, inspire you to practice, and provide individual, specific feedback is key. http://tashfeen.pbworks.com/f/EricssonDeliberatePracticePR93.pdf
Agreed. I had not read the Raj Chetty, that's a great citation.
Pleased to see discussion of the Oxford/Cambridge tutorial system. Particularly given the earlier start made on University in the 19th Century (and prior), I think it goes a long way to explaining the persistence of technically challenging and "elite" arts. It's mostly gone — no longer in STEM, and in a much reduced fashion in the humanities.
My sense is that the kind of talent needed to produce geniuses can be provided by aristocratic tutoring (in its different forms). But that same talent pool may be necessary, but not sufficient, for the actual production of genius.
At that point it seems like you need a confluence of historical and even species-level factors: there are "mute, inglorious Miltons" in many country churchyards, simply because the moment for that kind of poetry had passed.
The positive news is that (in my humble, but well-informed opinion) we are actually entering a new period where we will expect genius to flourish. In areas ranging from quantum computing and formal mathematics to political philosophy and cognitive science, we're entering a new period of low hanging fruit. Come join us!
I think the argument in your original post came across as more mono-causal than you intended, which seems to account for most of the disagreement between you and Scott.
Yeah, I agree, I'll take responsibility for that. But there's always this balance when writing these sort of pieces between (a) writing something that reads well, and (b) throwing in a bunch of caveats and hedges. If you do (b) too much you make yourself immune to criticism, but at the cost of being boring and making the pieces more longwinded and dragging. Frankly, I find a lot of online writers pick (b) and their stuff becomes. . . not a pleasure to read. Here, I probably strayed more toward (a) than (b), and in hindsight could have adjusted that ratio, but I also didn't know that Scott Alexander was going to pick over it.
Another reason your theory seems plausible (but would need more historical anthropology type work to back up) is that it would seem that tutoring would be a obvious educational style in societies over the past few hundred thousand years which didn't move into agriculture/hierarchy/class and therefore move people into specialization away from kids. If your way of teaching is 'do things with adults', which I expect most human societies have been, (i.e. learn what every plant and animal in an environment is, how they relate, and how to use them from elders) then it makes sense that you would have tons of geniuses - but before printing presses and writing and such, we wouldn't necessarily know about them, and their genius might be like "astonishingly good at hunting whales". Then with agriculture/inequality/etc, what used to be for everyone becomes limited to the rich - and the tutors got more specialized? Maybe? I've been reading Graeber (Debt, Bullshit Jobs, and Dawn of Everything) as well as a book called "Hunt, Gather, Parent", which I found very insightful besides the title, and which might have some good thoughts for this idea.
That's quite interesting. I've been making my way through the Graeber as well, perhaps I'll end up discussing it here.
Caplan did go on to homeschool his sons through high school as well:
https://www.econlib.org/our-homeschooling-odyssey/
Thank you - that's an update I was looking out for.
Also, "With the noble exception of their calculus teacher, my sons’ high school teachers just showed videos and treated teens like babies." - yikes!
Three points I want to bring up:
First, the empirical evidence in favor of elite tutoring over humdrum tutoring is not there.
That is, studies attempting to pit experienced against novice tutors repeatedly find no difference in learning outcomes.
The human tutors I've asked the last few months all agree with this. E.g.: "Yeah, there are some guys at my firm whom I don't think are very good, but their students' scores go up the same as everyone else's."
This is closely related to the interaction-plateau hypothesis in intelligent tutoring systems (very smart feedback-giving AIs only produce marginally better outcomes than fairly-dumb ones, so long as they're enough for the student to recognize their error), which is a major motivating drive behind the approach I'm taking for my own intelligent-tutoring work.
But, this can't be generalized too far: A typical study in this genre splits students into two groups, gives them a grad student who is either an experienced or novice physics tutor, and then compares their test scores. None of them look at 12 year-olds who have spent years with either a professor as a tutor vs. an undergrad, and see how many of them have started their own podcasts.
Second, apparently government requirements for mandatory learning are not as strong as people think they are, at least at the high school level, at least in some states. 100% of what I know about this I learned a few days ago from talking to the founder of https://www.powderhouse.org/ , who (slight caricature) told me that he can get students to satisfy all the state requirements for high school graduation by building some Github projects and then writing about it.
Third, I really want to see a more precise definition of genius by which we've declined.
I'll give an example of someone modern who was often called genius.
My friend, Michael B. Cohen, has published 15 papers in computer science since 2018.
This would be pretty good for anyone. But one makes this incredible is that he died in 2017 at age 25. He has over 30 publications total. https://scottaaronson.blog/?p=3468
From the stories I heard at his memorial, I'm pretty sure he attended normal public schools for K-8, and a public magnet school (Montgomery Blair) for high school.
---------
As a bonus, someone who did receive some elite tutoring in high school, and then went on to a great career: https://newcriterion.com/issues/2012/10/the-fifth-problem-math-anti-semitism-in-the-soviet-union
We love these write ups! Also, thank you for mentioning our work rethinking education at the CollinsInstitute.org
One compounding factor that makes this kind of theorizing difficult is that your data could be framed a different way: Genius is not created by private tutoring but instead stamped out by (only) going through the industrial schooling system. This could be solved if we had some other schooling model providing a large set of data points but we don’t. In history the vast majority of people were either industrially schooled, tutored, or industrially schooled and tutored on the side (with “tutored” often being a catch all term for a huge range of pedagogies that are delivered on a one to one basis). (Perhaps how common people who got both industrial schooling and tutoring are on your genius list means a better way to word this theory is, "people who's only interaction with an educator is through the factor educating process during their developmental stage have their chance at being a genius significantly retarded".)
Or at least we did not have any alternative models being tried historically. One thing that really excites us at the Collins Institute (outside of the unique pedagogy we developed) is the rise of radical unschooling (essentially just letting kids do whatever they want), not because we think that unschooling is the optimal educational pedagogy (we don’t) but because the data it is going to produce could change the way we think about producing geniuses. Essentially, radical unschooling will produce a true baseline in education—”what happens if we do not interfere with a child's education at all”. The early data here (which unfortunately does not control for socioeconomic data) seems to indicate that these kids are outcompeting students of the industrial school model—essentially meaning that the industrial school system is “worse than doing nothing”.
Also, as someone else mentioned, László Polgár is a person to look at in this space. The fact that he was able to attempt to create geniuses within a specific modality and do so three times out of three indicates a better educational system would not be 20% better but could be something like two hundred times better.
Note: To your question of whether the rich still practice tutoring. Yes, they do. When I was growing up in the 90s I had tutors come by once a week in a couple subjects to explore them at a level beyond what the school was teaching me. I know that tutoring was their full time job so there must have been at least a few other families paying for tutors in the Dallas area in the 90s.
It should be noted that standardized testing is the opposite of tutoring. It is quite literally a technology created to eliminate the need for one on one attention.
The point is that if a family adopts homeschooling mostly for ideological reasons it is likely it will just throw away the parts it objects to (say, Evolution and sex education), which CAN make the curriculum even worse and does not do much for the efficency of the proccess. Say whay you will, but it is not sex education (alone) which is holding American students back.
Now, if a family adopts homeschooling because they think it will produce more intelligent or more knowledgeable children, it is at least teying to do what Mr. Hotel is concerned about.