Wait a moment. I can agree that it’s a “common sense” guess that
P(bad spelling | less education) >> P(bad spelling)
but what that “technique” would imply is:
P(finds text with bad spelling belivable | less education) >> P(finds text with bad spelling belivable)
...which is kinda more interesting. So to speak.
If we’re rationalist, we don’t start our reasoning with “If we’re rationalist, then...”.
You are of course right about word usage, “learning” can various things, let’s say:
1) inspiration and motivation,
2) fuzzy background knowledge and a right mindset,
3) explicit knowledge of a theoretical framework,
4) exploring the framework—filling gaps, building bridges, internalizing concepts,
5) training (theory, offline, field),
6) getting community support and help with problems,
7) organizing and simplification.
But then if I consider each of these separately, the current LessWrong seems lukewarm in all but 2), and I can easily point to (for me) more effective ways to accomplish the same goal for 1), 3), 4), 5) and 6):
1) read thousands of pages of Eliezer’s braindump in a short span of time,
3) work through math and physics textbooks,
4) working together with people who are currently at a similar level, and are also learning about that particular topic (it does not matter if someone else did it before or not),
5) following individual people you can treat like your “jedi masters” (their methods and personalities need to resonate with you),
6) meeting folks in real life.
So I’m not denying 2) is important, but I think it rarely needs to be explicitly pursued, and also it comes with the danger of turning into a TED Talk-like disaster.
So maybe it would be a good idea to split/organise the content in LW depending on which part of “learning” its supposed to help with?
OK, so let’s look at the “ratio” problem: it only exists if you assume that every new addition lands in the same “pool” as the rest, right?
So the way to solve this would by introducing some organisation. Anything rigid probably wouldn’t work, but what could is maybe something I would call “organic organisation”. For example:
make it easy for each user to leave a “trail” of concepts, ideas, articles, links etc. that were useful to him at different stages of progress in a given topic,
make it easy to follow trails of other users, esp. those bookmarked in the past,
(maybe later) use the data for auto-organizing the content, but in a way that adapts as the community discovers new paths that lead to the same knowledge and skills.
This way you don’t have to say the same thing twice, you can say once what was your individual approach to a given topic, and only add a new item/page when you can’t find anything appropriate in the existing material.
Anyways in the meantime, the rationality blogs kinda work in this way, but they are “heavy” in the sense they require a lot of resources from one author, so there aren’t many of them.
Granted, but I don’t quite believe I have already extracted all the useful ideas from all the posts and comments so far published. It’s just that it’s too daunting a task to go find the rest of the good stuff, if it’s distributed too sparsely.
I may not have communicated the intention behind that sentence well enough. If you are exposed to an abstract idea A, and you have no use cases for it, it’s totally OK to invent a fictional scenario in which it could be useful, or other ideas it could go well with. I still think there needs to be something.
Hmm, sure my approach is definitely instrumental rationality-oriented, but I value epistemology a lot and you won’t find me complaining about it. As far as I can predict the experience of someone who has a pressing need to learn epistemic rationality efficiently and tries LW, they are going to be very frustrated (beyond the standard sequences). Eliezer worked not only as idea-adder, but also as idea-distiller and sequence-stringer. So maybe it’s just that the rest of LW engages in idea-adding only?
About my instrumental ideas, sure I’m interested in sharing them, but because of excessive lurking I have built quite some inferential distances in a few areas that are important to me. So for now I feel like it’s easier to write about things that I do not know too much about… (It’s actually a good meta-example of how “sentence-stringing” could be seen as the real “magic” behind teaching and learning rationality, and it’s (maybe?) a separate vital skill not many people have?) I’m generally baffled about how to communicate about this, especially the stuff related to “rationality of happiness”—I guess mostly because I know this part would sound utterly uninteresting. Mostly: here’s a bunch of methods that work not too bad, if you fine tune them for a long time… here’s some splitting of mental buckets to have more nuanced language… here’s a few tricks I stole from various sources and tested empirically… here’s my rough model of how to start success spirals of self change by slowly building confidence and accountability, but who the hell knows how it works really, I only tested this on myself so there may be dozens of other factors. You get the idea.
All this reminds me of how it typically goes when you try talk to people about regulating sleep. Problem 1: everyone is an expert. Problem 2: there’s no single method that works. Problem 3: no method works instantly. Problem 4: for anything to work, it needs to be fine tuned for the individual and it also depends on all other factors, so you can’t test these things in isolation. Problem 5: hearing a description of a method that works does not seem to justify the effort, until you experience by yourself what the benefits are. Problem 6: the benefits are spread over time, so it’s hard to notice them even if they are big and obvious in the “big picture” view.
All of this basically applies to teaching/learning instrumental rationality.
Yes, yes, yes, that’s a crucial insight about inferential distances right there. For rationality content that is useful to it’s readers, stringing items together is the most important part. Seemingly, no one is seriously doing this now (not even updating old paths to reflect better understanding or methods of transferring knowledge...).
I remember reading through that list sometime in the past, and I wanted to point something out to you.
[Disclaimer: all of the below is per my current understanding. It is a strong opinion moderately held.]
Sleep regulation is an example of optimizing a highly non-linear and volatile system with a multi-dimensional parameter space.
And in this class of problems, listing various parameters is good only as a way to know what is the space we are trying to optimize over. But if you try to gather information about how useful is each of those, you are shooting yourself in the foot before you even started.
If you hear a report of a method that worked for someone, it merely means it was the last missing piece to reach a local optimum.
In other words, this class of problems inherently do not have stable object level solutions.
Edit: please tell me if what I’m saying sounds wrong to your ears, I’m afraid I’ve forgotten myself a little and ignored the possible inferential distances I might have here and there. So from my perspective this simply points to the idea to apply and test some of the meta-level strategies that work in other contexts, like timeboxing imitations of various people, or upsetting the system on purpose to find a new local optimum, both of which may work better than random walk on the parameter space.
I guess this is analogous to “magic” as used in http://lesswrong.com/lw/ix/say_not_complexity/
Still, good post, and definitely closer to real life than Eliezer. Thanks!
But it seems to me that a good way to get “fuzzy” background context is to bang your head against a variety of concrete problems (rather than explicitly being fuzzy). Example: you don’t know how to approach a math problem, so you solve a lot simpler/related/example problems and it gives you a better “feel” and intuition about the objects involved. Then you come back to the original problem. I’ve seen this pattern a lot.
Can you give some examples of the opposite being true?
Sure, that’s the whole point—my claim is that to get to the stage where you do something actively, you need to pass through the stage where you grasp what an application of the thing would even look like. So I guess the post can be seen as pointing out it’s useful to notice when this intermediate step is missing, if you care about the quality of your learning. Note that it’s a heuristic that can be applied instantly and subconsciously, as opposed to “doing something actively” (which may require resources, waiting for external circumstances etc.).
It’s OK, since you agree on the “not being fuzzy explicitly” point, I don’t have anything more to say about it.
Don’t treat this “one sentence summary” thing too strictly—it’s kind of a reflex/shorthand that is useful in many situations, but not all. I like it because it’s simple enough that it’s installable as a subconscious reaction.
I think talking to this guy has gone past the point where it was amusing.
I propose to from now on, adhere to a fully general Against-People-Who-Don’t-Use-Letter-Case policy: ignore any post that contain’s only majuscules or only minuscules.
That should do well enough.
The commentator (on the Deepmind channel) calling out several of AlphaGo’s moves as conservative. Essentially, it would play an additional stone to settle or augment some group that he wouldn’t necessarily have played around. What I’m curious about is how much this reflects an attempt by AlphaGo to conserve computational resources. “I think move A is a 12 point swing, and move B is a 10 point swing, but move B narrows the search tree for future moves in a way that I think will net me at least 2 more points.”
If the search tree is narrowed, it is narrowed for both players, so why would it be a gain?
And looking at how he used up his time much sooner, he was more cautious today. He still lost and probably also took a psychological hit, so now my estimate of chances of Lee Sedol winning the whole match went down to ~5%.
Why isn’t it obvious?
I know what I’d do.
Run the algorithm on the Bitcoin market, and then on the stock market.
Upvoted for pushing EA to mainstream. I didn’t personally like the style of your article, but then I’m not the target audience of TIME either.
Question: does it benefit the cause to portrait EA as a “nascent” movement?
Of these three, only love bombing was prominently present, but I think it was actually genuine feeling in the “low rank” members (and I’ve talked to quite a few). Of uniforms, I think they all had unassuming trinkets/medalions hung from their necks. Food sharing was not present and the mentioned lunch was not more special than any other lunches I went to with other groups of people.
I think there were quite a few other interesting factors at play, and that’s why I suggested we could all learn something from a discussion about it. I don’t have anything like a complete analysis ready myself, but I detected significant amounts of:
casual, friendly touch—there’s nothing wrong with it and I would like to see more of it in other communities,
making good use of good looking people and sexual attractiveness—subtle enough to not be pretentious,
telling people to think for themselves (but I guess on meta-level, you already know the conclusion you are supposed to think of),
scripted/planned/predictable ways to evoke positive emotions,
honesty about emotions and subjective experience, and full acceptance of this by other members,
generally positive atmosphere of friendliness and helpfulness that carries over to the rest of the world, not just the in-group,
excellent stage/public speaking/PR skills of the top ranking members,
mixing in genuinely worthy/morally sound cases, like stopping wars (but from LW point of view it looks mostly like cheering than taking action).
Just mentioning, but It’s a good policy to avoid feeling good about figuring out anwers you already knew. ---> http://lesswrong.com/lw/il/hindsight_bias/