Strategy
The nine components of ultralearning, according to Scott Young:
Meta-learning: Learning how to learn the subject in question.
Focus: Cultivating the attention, environment, and schedule to get in the needed hours.
Directness: Learning the subject as directly and concretely as possible and avoiding “detours”
Drill: Attacking the weakest point in your knowledge with drills.
Retrieval: Forcing yourself to learn more by output than input. Test to learn.
Feedback.
Retention: Understand what you forget and why and learn how to remember.
Intuition: Cultivating a deep understanding.
Experimentation: Push your knowledge further through unguided exploration.
Though there do appear to be a lot of redundancies in Young’s Ultralearning concept, rather than try to refine it without having applied it in the real world before, we will be taking these nine as our strategic blueprint. We may change our strategy and make adjustments later, but for now we will take the advice of the ultralearning expert as directly as possible.
Metalearning
Young gives us some guidance on metalearning such as
Determining why you want to learn what you’re learning.
Observing and copying how experts or others have done your project before.
Devoting approximately 10% of your expected learning time on metalearning.
Determining what you can ignore.
So why learn the CTMU? Well, first off it’s fascinating. It’s in some sense a mathematical theory of religion which means it’s less likely to become dry and dull like so many other pursuits do once you dig into them. Further the CTMU purports to offer a sort of alternative foundation to science and mathematics, and if it lives up to that promise it could be very interesting indeed. So we are learning the CTMU for its own sake then, and not as an instrument for some other goal.
So who has done our project before? Who has learned the CTMU to both philosophical and mathematical mastery? There is not an abundance of people outside the author who have done this. If no one outside the author has done this before what might we do? Attempting to retrace the emotional and motivational steps that lead the author to write the CTMU as well as his major influences seems like a good start. In one interview I recall him saying part of the development of the CTMU was as a justification or argument defending his religious or spiritual experiences in his youth. He was also heavily influenced by John Archibald Wheeler, and has mentioned Thomas Aquinas and Wilhelm Gottfried Leibniz in interviews before. (I’ll dig up those specific interviews if I can, in the meantime this is a good one). So what can we do? We can start by going through the Wheeler papers cited in the CTMU.
This project is also somewhat “recursive”: in learning the CTMU there are going to be plenty of fields we are going to need a grasp of including Model Theory, Quantum Mechanics, General Relativity, and several others. For each of these there are almost certainly experts who can be copied. In a meta metalearning twist, we may even learn from Scott Young and his MIT challenge: at first he tried to do one class per week, but then to avoid the retention issues cramming is riddled with, he switched to doing multiple courses concurrently. The highly interdisciplinary nature of the CTMU also supports learning multiple ideas concurrently over a longer period of time rather than one by one as fast as possible, as learning them simultaneously may help integrate them.
Because we have multiple layers of learning we can also have multiple layers of metalearning, so our 10% of time dedicated to metalearning can be distributed to some extent and outside of the core knowledge need not be done all up front.
So what can be ignored? Since we mostly need to be able to use the mathematical knowledge, rather than proving it, we should be able to ignore a lot of the proofs in the mathematical texts, at least up front. This is very, very good news since proofs are the most expensive part of mathematical knowledge. In effect, we will be hoping we can rely on Applied Mathematics as opposed to Pure Mathematics. We likely don’t need an exhaustive understanding of each of the philosophies the CTMU draws from, merely a deep understanding of the core. We will also need a deeper understanding of some philosophers than others, for example Kant and Leibniz are likely much more important than Hume or Berkely. Physics is likely to be the trickiest background knowledge but lucky me, my undergrad was in Physics so I won’t be starting from scratch.
So what’s our overall metalearning start?
Understand Christopher Langan as a person and an intellectual. Understand and demonstrate how his influences and motivations could combine to give birth to a new theory.
Collect the important categories of influence and perform metalearning on them.
Collect the important categories of knowledge and perform metalearning on them.
Focus
This is likely to be a tricky one. I have a job, a wife, and a dog, interruptions are frequent and time is limited. I can block off larger chunks of time daily or almost daily, but not enough to be sufficient. That means we need two major strategies for focus. First, blocking off scheduled, lengthy, uninterrupted deep focus time. Second, finding ways to do the ultralearning in smaller, less predictable gaps of time. Let’s look at the second since the first seems fairly straightforward.
Flashcards or spaced repetition apps like Anki could help for the knowledge that is primarily association based, though certainly less well for things like executing mathematical algorithms. Microlearning is also a concept I’ve seen thrown around recently as a healthy replacement for doom scrolling. A cursory search for microlearning suggests it is “a teaching method that breaks down learning into small, focused units, usually under 20 minutes long”. 20 minutes is good, but shorter would be better. Ideally we should be able to address gaps of time as short as 20 seconds. Not that all gaps will be that short, just that for shorter sporadic sessions we want to be able to learn or practice something in durations that are as short as possible. Flashcards would be amazing for this, but again, not ideal for every kind of knowledge. We will get better as we go but we may also try using question cards for short, open ended retrieval tests, and derivation or proof worksheets of running mathematical reasoning that are formatted for convenience and can be worked on in short bursts and checked later. Having an AI like ChatGPT quiz me could also be a good approach to shorter learning intervals. These will all require significant preparation which is something to consider. A habit must also be built.
Burnout could also be an issue, I will address this by taking one day per week off.
Directness
This strategy of ultralearning overcomes the transfer problem of learning. How many of us have taken a courses in a foreign language for years and still been unable to speak it well? That’s the problem of transfer, things learned in an abstract context do not always transfer well to their concrete application. Learning a language by speaking with fluent speakers sidesteps this problem entirely. Directness is the solution to this problem, to learn by doing the thing you want to learn. Not to ignore the theory, but to learn that at its proper time as well.
The CTMU is a fairly abstract idea to begin with which may simplify directness. Math is also abstract however and one learns math by doing math. Unfortunately, unlike most mathematics textbooks, the CTMU does not possess problem sets, let alone an answer key to go with them. This simplifies how we can be direct, since the options are limited. We can learn the background knowledge, be it mathematical or philosophical, in the standard way of problem sets and essays, but also by applying those domains to the CTMU. We can learn CTMU concepts by applying them to the needed background knowledge and by developing proofs, derivations, or lines of thinking from the motivation behind the CTMU combined with its background knowledge to the CTMU concept itself (and ideally back again). This should help us learn the lay of the land directly, as though walking the paths themselves, rather than just studying a map.
Relating everything to the CTMU and studying the CTMU and multiple areas of background knowledge at once should also help us stay on task. Same thing with studying the motivations behind the CTMU.
Drill
I can’t think of much for this one up front without knowing what my weak points will be. We will improvise and learn as we go on this one.
Retrieval
Ultralearning strongly prefers active, over passive learning. We will always prefer to answer our own questions before referencing the book than trying to read the texts and remember them without working through each section ourselves. This is why for shorter learning blocks of Focus we suggested flashcards and worksheets, because they are active, rather than just listening to lectures, which is passive. As I read, listen, or watch I will also make quizes for myself with reference to where the answer can be found.
Feedback
For the CTMU directly the lack of problem sets with answer keys is problematic but this is not the case when it comes to other domains of knowledge. There are tests available for free online in many places. Additionally, I have friends who are willing to listen to me and give me feedback in the form of objections or requests for clarification on some topic I haven’t illuminated properly. This is indirect feedback since it’s more how well I can convince others, but I’ll refine as I go and find better feedback methods where available. Reformulating background knowledge problems in CTMU (meta)formalism and ensuring the same answer is given in the limiting case is one option that looks promising up front.
Retention
This, like drill, is going to be highly case by case basis.
Intuition
I found this principle a little more of an objective than a strategy, but we won’t ignore it. In addition to the Feynman technique, I will attempt to apply the concepts in question to everyday life, or think about them using symbols from everyday life, until an intuition develops.
Experiment
Our experiments will likely also serve as our demonstrations of understanding for the various concepts and background knowledge we will need. These will likely take the form of papers, essays, mathematical derivations, exams taken from online, and computer programs utilizing and illustrating the concepts.