Learning and the Subconscious | How to Become a Great Intellectual

This is a companion discussion topic for the original entry at https://www.youtube.com/watch?v=6BUmbi8OJqA

Notes written during the video:

  • Your subconscious has most of your brainpower
  • You need to train and use your subconscious to be really smart
  • Improve your subconscious with practice
  • There are mental skills, not just physical skills
  • Practice lets you automatize skills
  • This frees up conscious brainpower for learning new skills and then automatizing those. Then repeat to make ongoing progress
  • Automatizations can be changed later by de-automatizing, taking conscious control, making changes, then re-automatizing the changed skill
  • For mental skills, you need to come up with practice activities and do them
  • To use ideas/skills in your life, you need to learn both how and when, and you need to subconsciously automate most of the work
  • Your subconscious needs to recognize potential errors and flag them for conscious thinking. It should catch nearly all the errors (of particular types you learn about) while having few false alarms
  • The more you’ll build on or reuse a skill, or the more important it is, then the higher quality subconscious automatizations make sense – it’s more worthwhile to put work into it
1 Like

Current main skills related to CF that I’m consciously working on:

The grammar is something that I’m sure would be advantageous to automate. I’ve tried doing it while practising the other skills and I think it just makes me worse at both. In other words (referencing 18:40 in the video) I’m overloading my conscious thought bottleneck.

The other two I don’t think can be meaningfully automated. I guess e.g. if I knew Popper well enough to read his books and analyse them that would be a kind of automation. But I don’t think it’s a priority to know all of Popper’s writing in lots of detail, I’m interested in beneficial ideas that I can use without knowing all of the details too.

I think I can do well enough for now at studying and writing articles for the amount I’m doing it. I want to improve at it but don’t think it’s the right skill to develop consciously at the moment. I probably make quite a few mistakes in both. I think once I have a good automatised understanding of grammar I’ll find it easier to work on studying and writing skills (such as paragraph trees, reading comprehension, research skills). I think studying and writing articles at the moment can also help me identify specific areas I need to improve on, so it’s useful to do even knowing that I’m probably making mistakes.

Learning grammar is about changing automatisation for me (referencing 21.40 in video.) I already used a lot of grammar subsconsciously. I don’t have a good sense of how accurate my subconscious grammar is. I think I get enough right most of the time for people to mostly understand me, but I don’t have a way of measuring that. Even if I was super good subconsciously (which I don’t think I am), I still think it’s useful to understand the grammar language so I can talk about it.

The biggest change I think in learning grammar for me now is that I can go back and read over my sentences and more often than not describe which word is serving which purpose - though in more complex cases this is still a conscious effort for me. It has also made me more mindful of trying to write complete sentences; that’s something I’m currently consciously doing as I write and I’m rewriting sentences a lot. It would be better if I didn’t need to rewrite so much.

Maybe a sub-skill to develop for improving my use of grammar is having more complete thoughts before I start typing. I sometimes start typing and fairly often change my mind or add the rest of a thought mid sentence and that leads to lots of rewriting and possibly introduces more errors. I guess that would be an example of making more automatisations conscious.

Noting I’ve stopped watching at 25.03 for now (Mental Skills) so I can come back to this another time.

To what extent does repeated conscious analysis train the subscious? If you do conscious analysis on similar things again and again does it start to become a subsconsious skill?

Is there a specific step in practicing that makes skills subconscious or is it primarily about volume of repititions?

Is it mostly about finding things you can already do really really well and >99% subconsciously and then only training the subconscious on one very tiny skill at a time? In that scenario, are you doing a tiny bit of conscious analysis while mostly practicing in a more rote way?

I don’t know if it repeatedly practising something essentially develops subconscious skills. I think it’s possible to do the practise and not develop subconsciously, and there’s some extra process to automatise it.

I have a guess at this extra process:

I think of it a bit like compiling code. Programming languages have tons of options that are basically there to make it easy for a programmer to read the code (intuitive variable naming practises, comments, etc). None of them are necessary for the machine to actually execute the program and so it can be removed when it’s compiled. I think of transferring conscious skills to subconscious a bit like this, where the extra details (like all the words that you can explain a skill with) are stripped out of the skill and it’s streamlined to the functional components.

This doesn’t necessarily mean you can’t explain the skill any more, that knowledge could still exist separated off from the skill execution knowledge. By having the explanatory knowledge separated out it doesn’t get recalled when executing the skill so saves processing power and memory and allows skill execution to be more efficient.

I think this process typically happens subconsciously; I don’t know how it would be done consciously there may not be enough conscious processing power to do it. I think there’s some sort of subconscious process going on deciding how to make the skill work more efficiently (and presumably some sort of subconscious decision that making the skill more efficient like this is beneficial.)

Continuing from 25:03 where I left off.

27:40 coming up with practise activities.

I’m trying to finish working on some threads at the moment (watching this video and my C&R topic) so I can open some new ones (Peikoff grammar course and studying The Fountainhead) and not get so mentally cluttered.

This is itself a practise activity - working on my problem with overwhelm by focusing and finishing (or reaching a good stopping point) in some things so I can start others.

The grammar course and studying The Fountainhead are both practise activities.

I’ve been apprehensive about the grammar course because it’s long (20+ hours of video) and seems like a big time investment, and I don’t know how modular it is/how much I can just do it a piece at a time. I anticipate taking a few weeks to get through it all. Clearing some mental space settles my apprehension.

The Fountainhead is a practise activity I’ve chosen to practise grammar skills and reading skills. As well as exercising grammar trees (and later paragraph trees) I’ll do some more exercises where I write summaries and thoughts about it. I’ve chosen this book because I’ve read it a lot and like it and think I can fairly easily cross reference stuff from memory. I’m interested in writing some articles about it and may do so off the back of using it as a practise activity. Writing articles will also be a practise activity.

From the article (32:00 in video):

People usually only spend around 2-4 hours per day using their conscious mind heavily.

I intuitively agree with this, it feels about right. I don’t understand it though. I’m curious what the mechanism behind this is. I guess because it seems desirable to be able to change this and double or triple the figure. I wonder if it’s related to the mechanism behind sleeping.

35:00 (about when to automatise skills):
This part prompted me to think about conscious vs subconscious as two different resource pools as an approximation. Suppose you get 12 units of subconscious work and 4 units of conscious work per day (roughly one unit per hour). Then you could think of whether it’s worth practising automatisation in those terms.

So if a skill isn’t risky and you only plan to use it once or twice, you could estimate the number of conscious work units required to automatise it and compare that to the number of conscious work units to do it consciously. So if a skill takes 8 units to automatise but only takes 2 units to do consciously, it probably wouldn’t be worth automatising if you don’t expect to ever need it more than 3 times.

I think there’s a bit more to it than this - automatising a skill may have other benefits too like being reusable with adjustments, plus the subconscious units are valuable too (just not as scarce). Spending 8 conscious units automatising a skill will still require the cost of using subconscious units to enact the automatisation (suppose it’s still 2 units, just subconscious instead of conscious). If the automatisation also saves you some units of automatisation later on for a related skill then the automatisation expense is more of an investment than a cost.

It might be interesting trying to work this system out in practise. I think it would be very hard to get good at estimating units of conscious vs unconscious work required as I don’t think there’s a good objective measure of these things.

Taking a break at 36:52 in the video.

Thanks for replying.

Yeah, that makes some sense to me. It seems like you need good conceptual knowledge of something before you can start automatizing it.

Sometimes parts of explanatory knowledge get lost even after something is automatized. Then you’re left with just the procedural knowledge. I have experienced this.

After thinking about this stuff a bit more, I think I’m unclear on the concept of automatization. I know that my intuitions about learning has lots of holes but my intuitive/subjective feeling is that I learn best by building up background knowledge with lots of connections between ideas. Practice seems to work best for me as more of a diagnostic/testing tool. I think I have gotten more out of practice when I’m using it to see where I’m missing stuff. Then, I stop practicing and research about the missing piece of knowledge.

More questions:
What criteria can someone use to distinguish between automatizating and incremental gains in conceptual knowledge? Does that distinction matter? How many reps should it take to learn an incremental skill if you’re practicing effectively?

Just re-read this article: Curiosity – Learning to Mastery and Repetition
It has some ideas about how to practice and use repetition to learn.

Quote from article:

You can’t learn merely by repetition, you have to think about what will and won’t work. Repeating can’t figure out solutions and can’t do anything to find or correct errors.

Some of my examples are simpler because people should master some easier things before aiming for some harder ones. There has to be a progression.

In order to effectively think creatively about chess strategies, you can’t be too distracted by remembering how the pieces move. Practice does help automate one’s understanding of the piece movement rules. But practice isn’t just about repeating things, you think through what the rule for moving a piece is and figure out where it can go – it gets actual conscious attention when you’re learning it. You couldn’t just repeat correct piece movements without conscious attention, as a practice method, because you don’t know them well enough yet. (You could repeatedly move a rook back and forth between two adjacent squares, or something else simple, and thus make correct moves without thinking about it even though you don’t know the piece moves well, but you wouldn’t learn much by doing that, that’d be bad practice.)

It’s the same with everything else. Interesting, creative conscious thought is always building on many layers of thinking that were conscious in the past but no longer require conscious attention – that attention is now freed up for more advanced things.

What kind of conscious thought should be happening while practicing? I guess that this question is related to attention to detail. How do you determine relevant details when practicing? The answer to that would seem to depend on your goals. What are good goals to focus on when practicing?

Continuing from 36:52

At 44:38 Elliot talks about errors occurring in frequently used ideas which have other ideas built on them.

It reminded me of the “Five whys.” technique that some people use to understand problems. Rather than just asking something in the immediate context the idea is to ask deeper problems in an organisation that are the root cause of a recurring problem. I think this could be a good technique to look for lower-level ideas that are a root cause of a higher-level idea that is going wrong.

I’ve finished watching the video now. Overall I think it’s really good and does touch on a lot of common issues people have with doing hard things (like learning philosophy.) I’ve been working on reducing my project sizes (i.e. doing smaller chunks) for a while as starting projects that are too big is something I’ve made mistakes a lot with. This article helps build up a better (and more explicit) understanding of why that’s a good thing to do, as well as many important helpful concepts such as having a high success rate with ideas before building new ideas on them.

In practise I’m still stepping back slowly the magnitude of my learning steps and working to find the level I’m at a high confidence with.

It’s hard to tell why you’re contradicting CF or if you know that you are.

I didn’t know I was.

I think using meaningfully is misleading; a better word would be usefully but still the sentence would be an incomplete thought.

What I said contradicted CF by saying that those tasks can’t be automated. This is incorrect; I don’t think there is anything I can do consciously that I can’t also learn to do subconsciously.

What I had in mind was that I didn’t know how to automate the process in a way that would be the best use of my time studying Popper.

My next sentence tries to clarify it but gets it wrong.

Calling it a kind of automation, especially given the preceding mistaken sentence, seems weaselly. I can see how it could be read as also disagreeing with CF but in a backhanded way.

The sentence should also have specified subconscious analysis and I think is currently confusing. Writing the sentence again with these changes it would be:

I guess e.g. if I knew Popper well enough to read his books and analyse them subconsciously that would be automation.

CF’s position is that automatization is a useful and necessary part of all types of learning.

Learning about Popper’s ideas from his books is a typical case where, as usual, you need to learn some small chunks, automatize most of that content, learn more chunks (more initial stuff or building on prior chunks), automatize most of that content, and repeat. Layers of knowledge should be ~fully automatized by the time you’re building several layers past them (how many “several” could reasonably be varies considerably by layer size – with really tiny layers it could be more than ten).

Automatized = subconsciousized = your subconscious can use this knowledge.

When your subsconcious can do something but with some conscious monitoring and guidance too (e.g. driving), I suspect that means a ton of chunks are fully automated and some are either not automated (subconscious can’t do them) or not automated well enough to use (too risky to let your subconscious control that part of driving).

1 Like

Thinking of driving as an example, I wonder if there’s a better way people could learn to drive that takes CF into account. I think typically when people are learning to drive they’re taking on a bunch of things at once and having to commit a significant amount of conscious effort to each. They’re managing the accelerator/brake, steering, indicators, lights, checking side and rear mirrors, reading road signs, reading road markings, observing traffic/pedestrians/hazards and (though not as much as earlier years) managing gear shifts. Learner drivers might study some of this at a theoretical level but I doubt many of them have much of it automatised before they’re actually driving.

If a learner could learn these components individually (e.g. with simulators/video games) it may reduce the number of things they’re consciously having to think about when they get behind the wheel of a real vehicle and pose a potential risk to life.

Good instructors try to account for some of this such as taking learner drivers to quiet areas where there is little traffic or confusing road marking. In this way mistakes like not checking blind spots or not using indicators have a much lower (but still non-zero) risk of disaster. The instructor may also be able to intervene if the vehicle has dual controls, and can in any case warn the learner if they spot a disastrous error.

I suspect my experience playing driving and shooting games (and the spatial awareness they require) may have helped me when I first learned to drive. I think there are a lot of other skills a person could already have automatised that would contribute.

I consciously remembered right after writing my previous post that there is such a thing as learner driving simulators and it’s weird that I didn’t integrate that properly.

I guess what I’m wondering is something more CF-tuned, focused on building specific skills that are part of driving in a piecemeal way.