# Project outlier (maybe more like experiment outlier)

The idea: I want to become incredibly skilled in a vast array of areas. As an example, I might benefit from making less crappy designs.

Will it work? I have absolutely no idea.

I’ve been reading atomic habits. While I think it’s not as good an exposition as it could have been (gotta hit that word count range if you want to publish a book!), and I dont think a fair bit of it is necessarily true…

one point I got from it seems almost like a mathematical tautology but also extremely powerful.

Imagine for a second that I’m a N-dimensional space with each dimension limited to the (0, 1) range, each describing a particular skill or knowledge or experience or behavior that I care about - 0 is worst and 1 is best.

I can trivially improve a component by 0.1% every day. Expanding my trumpet range by 0.1%?

That’s 0.025 semi tones. I can do that with 10 minutes of exercise.

Improving my fitness by 0.1%? A literal 10 minutes of walking are more than enough.

Improving my Japanese vocabulary knowledge by 0.1%? I already have stats, and with suboptimal parameters I can definitely do it in a day.

The book talks about 1% improvements, but the math is still powerful even if the improvements are an order of magnitude less.

Let’s define f(I) as the sum of the N components divided by N. Then trivially, if I improve something even by 0.01% every day, as long as i dont have on average worsening days, f(I) is going to be kind of monotonically increasing over my life.

I can read 5 pages of a psych 101 book tomorrow - I’ll be very slightly more knowledgeable about something, and f(I) will go up.

I can even estimate f(I) by self grading myself on a number of dimensions and seeing how the score evolves over time.

It seems a mathematical inevitability that if I apply an iterative approach to small improvements every day instead of following goals or targets, no matter how small the improvements, my score is going to tend to 1, no matter where i start from. Sure, i only have another 70 years to live if I’m lucky. But 1 would mean being the greatest polymath in the world and I dont need to go that far.

If i improve my vocabulary by 5 words a day, which believe me, is trivial and takes minutes to do, I can EASILY die fluent in 10 languages.

If I improve my range by half a semi tone per month for a while, I’ll be screaming triple Cs in a matter of a few years.

If I improve my staccato speed and cleanliness by 0.05% a day, I’ll be able to play the first brandt trumpet concerto in a similar amount of time.

If I write 500 words a day (takes me 20 minutes), I’ll have written several thick fantasy book every year.

Sounds too good to be true. But I cant see how it could be false.

I also have programming as a tool most of the population doesn’t have. I’m now taking private japanese lessons, and the incredible thing is I know almost every single word that comes up. Why? Because I was able to fine tune a language learning SRS to my strength and weaknesses.

I thought of how to further maximize the impact of time spent using a better heuristic for scheduling work and visualizing what needs to be done.

You cant do that on duolingo.

I plan to remove beeminder from my productivity systems - I think I have an idea that would better motivate me, and tomorrow I’ll code web and android apps.

Coding is like a superpower for this project. Want to make youtube less enjoyable? Randomly pause videos once every 5 minutes, show a picture of someone I hate laughing at me, then resume. I can still watch all the youtube I want. I just have to also take the loss from someone I want to “beat” rightly laughing at my waste of time while they’re more successful than me.

That ought to make it slightly less enjoyable. Or maybe not! Who cares. I can code it in 15 minutes and if it doesn’t work I can try something else.

Superpower.

Speaking of code, why does my autocorrect not correct cant? Is “cant” an actual word?

I’ll better organize my thoughts and post progress here.

2 Likes

Cant is an actual word. I think it means a private language.

3 Likes

For many goals, this is going to work: write 500 words a day, you’ll have written 500 x N at the end of N days.

But surely there are two problems you’ll surely run into with growth goals: one is physical limits, the other is forgetting.

• I can’t get 1% stronger every day forever. Not physically possible. (I’m reminded of the stories about a young man given a calf to carry about every day. As the calf grows each day by a little, he gets a little bit stronger. But no way is he carrying 850kg of bull around at the end of it all.) There are even weighlifting training regimens that go this way - Candito’s Linear Progression springs to mind. But nobody expects you to keep going forever.

• And this is just as true for mental acuity. I’m just not sure I can get my head around (say) quantum mechanics, no matter how many days I read another 5 pages. I’m just not smart enough.

• On forgetting, surely the risk is I can’t learn more vocab each day, without forgetting old stuff. I can learn 5 new words, sure - but do I still remember the 5 I learnt 200 days ago? You will have to continue to re-train (through SRS or whatever) all the old stuff, at the very least, which will take incrementally more time. I used to speak Swahili when I was young: I don’t now, because I never practiced after leaving Africa. I know that I’ve forgotten half the stuff I learned at college about group theory - I recently had to re-learn it to understand the group theory behind QM.

You will have thought of both of these objections yourself, I’m sure. And having said all of which, I’m also sure, with enough dedication, the limits are way higher than we let ourselves believe. Could you die fluent in ten languages? Almost certainly, if you really put your mind to it. Could you play your instrument to a professional level? Ditto. Write 3 novels a year? Sure. Code like a superman? You could get a long way. Could you do all of those at once? Maybe not enough time.

4 Likes

Thanks for the feedback! It’s useful. Don’t worry about bursting bubbles; if they’re really bubbles the need to be burst.

I have two comments here: one, I talk about 0.1% and in some cases 0.05%. I think that can work for a looooong time. Additionally, the first few weightlifting workouts you do increase your strength by double digit percent. But that’s not the problem with this reasoning - if I get to carrying a 250kg bull, I’m in the 99th percentile. I achieved my outlier status in weightlifting. Sure, that’s nothing compared to what strongmen can do, but they’re not in the top percentile, they’re in the top 0.0001% or something. Top 1% is still outlier.

I think you can. You just won’t start from quantum mechanics. Start from 5 pages of precalculus and master that. Then do 5 pages of basic high school calculus, then 5 pages of more rigorous real analysis, then 5 pages of complex analysis, then linear algebra, then physics 101, and so on. You can do about 3 books a year even just working through 5 pages a day. 2 pages a day are enough to do more thna 1 book a year. And if you go through 30 books, I think it’s a reasonable assumption you can get to truly understanding quantum mechanics.

I have simulated (and experienced) this. Through SRS, you don’t really forget that much. And each word you learn will take exponentially less time to review every time you do review it. I already have words with intervals >1 year, and I do remember them, and the interval gets bumped to >2 years.

5 words a day imply something like a couple of hundred reviews a day in the long run, which is really nothing. They don’t really increase from there no matter how many new words you learn.

SRS implies practice each day, even though it’s very little.

Was it as hard as the first time you’ve learnt group theory? If not, you can view the previous experience of learning group theory not as getting to, say, 0.3 on the group-theory-knowledge-dimension, but to 0.1 which represents the knowledge that you’ll retain even after forgetting most of it.

I think I could die fluent in ten languages without putting my mind to it. 5 words a day in a SRS systems are really, really trivial. In fact, I might make an experimental SRS to use on the side that forces me to limit myself to 5 a day - maybe starting with spanish. I’ll show in a while what the daily reviews are like.

50 years, with 5 hours dedicated to getting better every day, are over 90 thousands hours. I think that’s plenty enough time.

2 Likes

Love this stuff! Thanks for putting this out here.

One thing I think you may be leaving out is making sure your daily practice is always a stretch. If you only focus on time and consistency, the amount you improve each day is likely to decrease over time, since what challenged you at the beginning is not likely to challenge you a month or maybe even a week down the road. Practice is only efficient when it pushes you to the edge of your current capabilities and gives you concrete, actionable feedback. A good spaced repetition system has both of these things built in, which is why I think you’ve experienced so much success with your vocab memorization.

I’d suggest reading Deep Work by Cal Newport. I think you’d find it to be super relevant to what you’re trying to do.

I’d like to quote little bits and reply piece by piece but to be honest I just 100% agree with everything you’ve said Considering my fitness lever right now, a 10 minute walk is going to be >0.1% of increase in fitness for a relativaly long time. But once it becomes trivial, to still make walking fit in my framework, I’ll have to switch to brisk walking. When that becomes trivial, I might switch to my indoor bike trainer, where I can increase and accurately measure the increase in watts I put out during a fixed time. Say, a 10 minute time budget - my smart trainer will report the actual watt-hours I put in by pushing on my pedals. Considering I’ll probably start with a 50W output or something, a 0.1% improvement there is, well, 50 milliwatts. Trivial! I could have a calendar with how many watts I should be putting out day by day.

I will definitely do that, then.

I’m working through atomic habits taking notes categorized on:

1. What I think is objectively true
2. What I think is very likely to be true
3. What I think is possible to achieve based on what has been achieved by other people.

These actually don’t come straight from the book. I might completely disagree with something the book says, but even then, I might not discover what I think is an objectively true fact if I don’t have that sentence to disagree with in the first place.

I’ll do the same thing with deep work. I also plan to post my rough conclusions in this format taken from Atomic Habits, so fellow bees can critique them.

1 Like

Thanks for the positive response to my slightly nit-picky feedback, @eugeniobruno

There is a question of capability limits here, though, regardless of whether you’re getting there 1% or 0.1% at a time. My reference point in this is learning mathematics: I have postgrad-level capability in some areas, and distinctly middling-undergrad level in others, and I don’t think it’s just down to practice and application. I feel there are definite, individual, limits on capability, by area (calculus vs statistics vs group theory, or whatever). I would to see this all the time in teaching environments. Some people would breeze through a particular field until they hit a particular concept, and then, no matter how hard they tried, it stopped being easy, and became hard, and rapidly became impossible.

Maybe this is a matter of willpower, but it seemed not to be. They were super motivated, but just couldn’t “get” the concepts beyond a certain point.

I don’t know enough about languages, so I don’t know how this problem of hitting your ceiling might map. It may be the concept doesn’t apply, as regardless of the language, there are millions of normal people who can speak it somewhere, which isn’t true of (say) advanced mathematics.

I love the challenge you’re giving yourself, though, so you get to be my test case in this field In the end it’s significantly about motivation, for sure (here speaks someone who really wants to play the piano well, but struggles to get his hand/eye coordination to move fast enough - maybe it’s practice, maybe I’ve just hit my capability limit - and yes, I’ve tried beeminding it!)

1 Like

But still, “middling-undergrad” level in a math subject is still an outlier when compared to the general population, right? If you have a “middling-undergrad” understanding of group theory, how much group theory knowledge do you have over 99% of the population?

I don’t think this is either. If it is about willpower, I am done. I’m never going to be able to do this. I’m basing this supposing it is on systematic behaviour change without significant increase in required willpower.

As an example, one of the first things I’m planning to implement - however stupid it might turn out to be - is to have my SRS pop up when I open a youtube video that is not in my music playlist I use for studying, closing and unpausing the video after 10 repetitions. This does not require willpower. It can decrease my time wasted on youtube, or it can increase substantially the amount of repetitions I do.

Good point. Also, the more languages you learn - generally - the easier it is to learn the next one. I’m Italian; I’m obviously going to have an easier time with romance languages than someone who only speaks English. If I also learn Spanish, French will become even easier than it would have been if I didn’t.

Again I really hope it is not about motivation, or innate ability, or predisposition, or any of that. I hope it is just a problem of optimizing myself just like I’d optimize a piece of code. Find the lowest hanging fruits, optimize those, repeat profiling, find the next area to improve… whether it’s changing a data structure or the way I organize my day!

I see, so this is about producing a process that does this for you, analogously to the way Beeminder helps us with akrasia. Which is very interesting, as it’s quite likely that at least some - and quite possibly a lot - of what works for you will work for other people. So if you succeed in your endeavour, you’ll have at least the beginnings of a tested general person-optimisation process. Pretty neat! I look forward to seeing how you progress!

1 Like

Myself, I look forward to see if I progress.

I’ll keep the thread updated from time to time… unless this fails horribly.

It’s too late now - you’ve preregistered your experiment with us, so you have to report the results either way, or we’ll think you’ve been p-hacking

3 Likes

Ah, thats actually unironically a good idea. Not to clear my name but because itd be kind of like p-hacking and the only one losing is me! Having a record of what I tried, what worked and how many times I’m wrong about the hypotheses I make is valuable.

3 Likes

I love this idea, and am hoping to do something similar in my own life; less trying to be an outlier and more wanting to be ‘good enough’ at a wide variety of things…

Trying to have a growth mindset, believing in the journey/process over the end result and be consistent seem to be the main things!

2 Likes

Documenting my thoughts on this.

But let’s reframe this for a second. I already said I don’t like the book, but even if I don’t like it, I can extract beliefs about what I think is objectively (or likely) true by reading about things I didn’t necessarily ever think about.

I’ve already said I have ideas that I think are pretty good and make a lot of sense and are fairly easy to code.

I’ll trim my list of conclusions from the book even more, make a TODO of things to implement based on the concepts I discussed by myself and with other people in this thread, and start to track how everything goes.

Ugh. I hate the feeling I’m feeling right now. I even know that it is not rational.

I think being good enough at a wide variety of things would make you an outlier. We’re not trying to become the best polymath on the planet here.

2 Likes

I love this post, and find it really inspiring and hopeful for myself.

I don’t think this is trivial at all. Rather, I think the only components you can improve by that much a day (yes, that’s a lot) are the very weak ones.

Why do I say that’s a lot? Because 1.001 ^ 365 = about 1.44. So that’s 44% better every year.

Seems to me at certain points you’ll hit diminishing returns and plateau, and getting past that plateau is going to take additional time and resources.

If you’re way out of shape, 10 minutes a day of walking can improve your fitness by 0.1% or more, sure. Once your body has adapted to the point where 10 minutes a day is no big deal because it’s used to that exercise, then you won’t get any additional fitness improvement - you’ll just be maintaining at best. So you have to increase the time or effort - start running, carry a heavy backpack with you, or whatever. Then that will hit its limits as well, and figuring out how to get past each plateau will take increasing amounts of focus, time, energy, and creativity.

There’s also an issue of “switching” - related to the “Deep Work” concept that @narthur raised. Reading 5 pages of Psych 101 a day isn’t going to be that effective - it takes a few minutes to get into the zone, a few minutes to recall where you were, a few minutes to add the new knowledge and integrate it with what you already know. You can’t cross a chasm in two small jumps.

You’re also going to run into Goodhart’s law - you’re not actually optimizing what you care about. If you read 5 pages of Psych 101 or learn 5 random words, are you actually gaining useful knowledge out of that? Like are you practicing using those words in context, coming up with them spontaneously, and so forth, or are you just improving your Anki score with no real-life gain?

I think what’s missing here is the idea of focus and intensity - becoming incredibly skilled in one thing is going to require laser-like focus and significant time and deep work.

3 Likes

No offense but why is your metric mostly based on being better than a % of the population? Isn’t that a bit less about improving yourself and more about the social sides of things?

2 Likes

Is knowing how to calculate derivatives and integrals a great thing to aim for? Well, if you live in 2000BC, that would be kind of unrealistic, with calculus not being invented and such. If you live in 2020 you were probably taught calculus in high school.

Is hitting a double C a good target on the trumpet? Who knows, if I don’t listen to and admire other trumpet players? But if Ilook at big band trumpet leads I see that they should be able to hit high F or high G reliably, so a double C is an ambitious but doable target.

How can you rate yourself without taking into account the rest of the world when you, in fact, live in that world?

By the way, before really starting with this I’m focusing on a gig that I got from a small local business, which is why I didn’t reply to the other two counterpoints yet.

3 Likes