There’s a problem.
Science is slowing down. Sort of. There are nuances.
But when you look at the overall pace of scientific achievement, what’s less in doubt is that it takes an exponential increase in researchers to produce a linear increase in scientific achievement.
This is problematic. Scientific achievement is important for a large number of reasons I won’t argue about here. We want science to move faster, not slower, or at the very least we want it not to slow down.
How can we achieve this?
The first option, in many ways the default, is to keep pouring more resources into scientific investigation. More money, more time, more people.
The problem with that is that we’re running out of people. World population is very close to peaking. In many developed nations it’s already declining. If we can’t support linear progress in science with an exponentially increasing amount of researchers, then absent another solution, science will slow down.
The second option, in many ways also a default (and something I’m quite hopeful about) is to use AI as a substitute for researchers. Perhaps exponentially-increasing artificial intelligence will substitute for flatlining human intelligence?
But there is a third option, and it in many ways is linked in my mind to why science keeps getting harder.
Yes, the low-hanging fruit get picked early, and yes, you can’t learn new things about physics from moving rocks through a wire anymore. But one of the biggest reasons I think science is slowing down is because it takes more and more time for a researcher to get to the frontier before they can advance it.
Time To The Frontier
How long does it take, as a student, to make it to the frontiers of a scientific field?
How many years of study does one have to go through to be able to meaningfully advance the state of the art?
In the current (American) system, students spend about 12 years - from ages 6 to 18 - in primary and secondary education. This doesn’t get anyone close to the frontiers of a field; in many cases, even for students and schools focused on science, it’s barely enough to cover the basics.
Then there’s 4 years spent as a undergrad, where a student may participate in research, but such students are rarely advancing the field themselves. They’re assisting other researchers, and learning how to do research, not actually discovering new things about the universe themselves.
That’s 18-22.
Then there’s graduate school. Assuming a student wants to be a researcher, they’re generally going to be getting a PhD, which takes somewhere around 4-7 years. A dissertation (which is usually required to get a PhD) is usually some bit of original research, so if we stop the clock there, a human being will be capable of original scientific research between 26-29 years of age, after 20-23 years of education.
That’s not the end of the story, however. Certain disciplines require postdoctoral study, and sub-fields of sub-fields might require even more time spent learning.
Ageing Achievements
There’s another line of evidence, masterfully collected and summed up here, that the age at which important research (e.g. Nobel prize-winning or similar) is done has been increasing faster than demographics alone would suggest. Across the 20th century, the median age of first-time mathematics and economics paper authors increased, along with the age of Nobel-prize winners.
Again, this suggests - not proves, evidence isn’t proof, but suggests - that on average there’s more knowledge one has to imbibe before one can meaningfully contribute to the scientific literature.
There’s also the idea that human brains perform worse in old age (while plasticity doesn’t stop once one is past one’s prime, it’s both obvious to observation and empirically valid that brains work less well in one’s old age) - meaning that the longer it takes someone to learn all the knowledge necessary to advance a field, the less time and brain power they have available to them once there.
What To Do About It?
If you believe, as I do, that the increasing time it takes to learn enough to meaningfully contribute to a field is a factor in why science is slowing down, then what can be done about it?
Granted, it certainly isn’t the only factor, and we should try to take all paths to victory instead of just one - but what does solving this particular problem look like?
Art is Long, Life is Short
In Scott Alexander's (of ACX fame) short story Ars Longa, Vita Brevis, he presents a society of alchemists with exactly this problem. (Spoilers ahead; go read it if you can. It’s not long, and very good!)
The alchemists are trying to create the Elixir, a potion that indefinitely extends the lifespan of the drinker. They are trying to make themselves immortal.
The problem they face is that, in order to create this potion, one must have so much knowledge in their head that they can’t possibly learn it all in a single lifespan. In order to create the Elixir, they must already be immortal - a Catch 22.
To solve the problem, the alchemists are divided up into four "classes." (I’m paraphrasing here.)
First are the alchemists who do the actual work/research, then write books on it.
Second are the editors who take those books and distill them down, making them as easy and intuitive and fast to learn as possible.
Third are the teachers who teach the alchemists as efficiently as they can, using the materials from the editors.
Last are the alchemists who aren’t all that useful in the effort to create the Elixir, so they’re stuck dealing with all the mundane problems of the world so the rest of the alchemists don’t have to.
Each class of alchemist exists within a feedback loop, iterating upon itself to get better over time. An alchemist charged with doing actual work is accelerated as fast as possible to the frontier, does as much work as they can, then writes it down. An editor gets better at editing over time, and books are written about the art of editing, such that future generations are better editors than previous ones. And teachers likewise improve their craft, accelerating their pupils faster and faster as they teach more and more effectively.
Thus the alchemists attempt to solve the Catch-22 by coalescing all the efforts of every alchemist until eventually, one of them is smart enough and has learned fast enough to get to the frontier with enough time to discover, and write down, the formula for the Elixir.
Could We Teach Better?
Scott’s Ars Longa, Vita Brevis presents a possible solution to the problem of an expanding scientific frontier:
Make learning more efficient, so that students can learn more in the same amount of time.
That way, while the knowledge necessary to reach the frontier keeps growing, the time it takes to reach the frontier stays the same (or at least grows slower).
And in the story, it’s the job of the editors to make teaching more efficient by making knowledge easier to learn. They find the best metaphors, the most effective teaching methodologies, the right problem sets, and so on, so that the teachers can convey the same amount of knowledge in decreasing amounts of time.
Imagine if this was applied to, say, medical school. Currently, in the US, new doctors tend to be in their early to mid 30s once they’ve completed their residency and any specialization.
What if they didn’t have to be?
What if, instead of the 11-16 years it takes to become a doctor from the end of high school, it only took eight? Seven? Five?
How many more people would want to become doctors? How many more doctors would we get, and how would that drive the price of healthcare down? How much money would be saved in student loans?
This isn’t a pipe dream; it’s a barely-explored path to improving a very dysfunctional part of our society. After all, schools are hardly incentivized to decrease the time you spend at them, especially when you’re paying them for the privilege. Why would they spend time and money on research that would reduce the amount of years you’re paying them?
And for those who think it’s impossible to achieve any significant gains by teaching more effectively: why not? There are clearly better and worse teachers, and better and worse methods of teaching. Plenty of famous geniuses had one-on-one aristocratic tutors in their childhoods. Why can’t this be extrapolated and optimized?
(There’s an objection here that goes something like, “But different people learn differently; there is no one Optimal Method of teaching!” To which I reply: then the solution is to have a variety of different options for different people available, like a learning buffet, such that each student has available enough material that makes sense to them to learn quickly. In the end, no objection really touches my central point, which is: teaching is not something that is currently done systematically or with the goal of optimizing it. There is a large amount of improvement possible on this front.)
So where are all the editors in real life? Is there anyone working on making existing knowledge easier and faster to learn?
Where Are All The Editors?
In our world, we've got scientists who do the actual work, and then write books on it.
We've got teachers who teach the next generation of scientists, who are generally scientists themselves.
But we seem to lack editors - people whose sole job it is to take existing knowledge and compact it into the most intuitive, memorable, and easy-to-learn format.
Are there such people?
One could argue that pop-science authors like Michio Kaku are doing some of the work here, distilling science down for a general audience, but I’d reply that the purpose of their work is at least as much entertainment as it is education. They’re trying to instill a little bit of extra scientific knowledge in the general public and inspire the next generation of scientists - good and necessary work! - but they’re not distilling and compressing existing science textbooks down so that students can absorb the same material in half the time.
Then there’s the existing fields of pedagogy and andragogy (the education of children and adults, respectively). Isn’t this what they’re for?
Well, not exactly. What we’re really looking at is the field of educational research, which is supposed to be about how to make education better. But this too is not really what we’re looking for, for two reasons:
It’s not a very successful field; aristocratic tutoring and phonics are its two most reliable results that I’m aware of
It studies ways to make education in general better, whereas we’re looking for the specific refinement of material to make learning it faster
So are there any Editors?
I think so, although they’re widely dispersed and ununified as a discipline.
First we have Kurzgesagt and other YouTube educators. They’re doing some of the work we’re interested in - reformulating existing material into a new medium - but as far as I know they’re not testing their methods to achieve faster learning velocities. I have heard success stories of innovative videos and new metaphors teaching difficult concepts in less time though, so let’s not count them out.
Next we have Numinous's Tools For Transformative Thought. By incorporating Anki into didactic essays, they create content that is literally more memorable.
(For those unaware, Anki is a flash card system that uses spaced repetition. Basically, some psychologists figured out about how long it takes the average person to forget something, and Anki is flashcard software that brings flashcards up for review right before you’d forget them. It is an amazing tool for memorization.)
After that, the only thing I can think of is the recent GOAT by Tyler Cowen, a “generative book” which allows readers to literally have a ‘conversation’ with the book via AI. Users can ask questions as they read, generate summaries, and so on. It’s a fascinating project.
Conclusion
Much the same way that modern food can be much more calorie-dense than historic food (and thus we can now consume far more calories in far shorter amounts of time), we should make our teaching materials more thought-dense, so that we can learn the same concepts in far shorter amounts of time.
While the calorie-density of food likely contributes to obesity, increasing the thought-density of educational material could make us fat in knowledge, which is a good thing (along with being a weird metaphor).
Having material that students can learn more quickly could be transformative to many areas of society, yet I see few people working on this idea.
Do you know of anything that qualifies? If you find something, let me know, and I’ll add it to the list.