It's at most half done, and it's probably confusing to read.
It might be a good idea to come back later.
Alternatively, you can prod me to write it.
Raising The Sanity Waterline by Growing Lesswrong by Making More Grayspaces
- idea how lesswrong could improve
- call-to-action on supporting me doing it
- tell me whether they’d be interested in participating on the things that could need participation
proceed:
I think the answer is to build a grayspace
Karma only seeps out
only after x karma things become lesswrong
people doing things on other platforms csn collect karma
- like point out things to other people when they’re wrong
codename dojo
people writing stuff and getting recommended the articles that already tackle that
- actually, can AI do that?
AI overlay (Jev) that shows which sentences might refer to which concept on lesswrong
a better explorable lesswrong wiki
people have a learning platform baked in that shows them how to write good posts or good comments, and can help each other improve in the dojo
start:
Lesswrong is currently in an interesting position:
- big (potential) demand from the world what’s going on + big hope of us that this demand gets met in a way that makes the discussion better
We (I’m going to include myself here even though I’m relatively new. Basically the only point in my favor is that I had updated on AI becoming a big thing after reading about it in 2018) have predicted all of this AI stuff was about to happen, and have been the place where the discussion for it got shaped.
Now that AI is a thing, and especially over the last few weeks things have blown up, we could make it so (and it probably already is that case) that Lesswrong is in demand.
At the same time, this is now our time. All this started ages ago with Eliezer wanting others that thought well about AI, and therefore writing up his thinking for us so that we could think well (about AI), too.
and then trying to talk to people about it, and then struggling tremendously. This made him make it so that he could talk to other people about those insights in a productive fashion.
And I think for a very small part of the population, we’ve succeeded. Instead of one Eliezer from 2006 that could think well about AI we’ve now got probably something like ~100001 people who can think well about AI. That is both an impressive growth rate, and an impressive number—but just looking at the current situation, we’re too late. Our time is probably close to being up, and we’re not winning.
But at the same time, the time for using our art has come, and the biggest avenue for winning is actually public opinion forming—which supposedly works way better when people are rationalists.2
So a straightforward way to solve this would be by just getting much, much more people acquainted with the readings, the community and the thinking.
Preaching to the choir here, I think we should try to grow the number of people who’re good at thinking about AI to probably something like 1% of the world population in the shortest time possible, and I’m happy to take any marginal improvements along the way.
So, let’s say we’re up to the challenge. How should we tackle this?
I propose by creating more grayspaces of basically any kind.
Over the last week, I contributed to one such effort (the creation of https://AGI.FYI ), but I propose we do way, way more of these.
Even more, it’s probably a good idea to view our whole endeavour as a funnel towards rationality, and optimize the whole thing.
§The funnel right now
people read HPMOR, ACX
people interact on twitter
people who read iabied
just random people on twitter, actually
llms tell them to post here
people see lw posts on hn or anywhere else
CFAR
people go onto lesswrong, read stuff there
comment there
and write posts there
interact with people there, and hopefully get good? Win?
And then there’s more adjacent things, but I’m not sure how much they actually help people have a better discussion culture:
- Clearer Thinking
§The goal: Completing this funnel
Caveat: This is me idea-generating, and not (greatly) limiting myself—please be annoyed about how someone could say something so stupid, then think for five minutes why my idea obviously can’t work, and then tell me.
spaced repetition on lesswrong
‘learning how to comment’ on Lesswrong + worked example + practice tests + near misses + spaced repetition
‘learning how to post’ on lesswrong + worked example + practice tests + near misses + spaced repetition
indexing all lesswrong articles with an concepts overlay where one can see what things other things draw from.
a third kind of forum (the dojo) around lesswrong (but that looks distinct!) where people can just start writing stuff, and then critique with comments, both factually and rationality-wise
- important: do we want to up the barrier for people posting things on ‘normal’ lesswrong?
- karma on the dojo comes from votes from ‘real’ users: e.g. when people have resolved a problem with their article in a correct way, and so on
Editor:
a writing guide fact-checking thing (Jev-like?) that tells you which problem(s) your sentences have
- could people do that and get karma for it? (not from the system obviously)
an agent that reads your text and suggests articles on lesswrong you might want to look into for writing this thing.
- could people do that and get karma for it? (yes, I guess?)
people get some kind of points for correctly calling things out to other people online?
people get points for referencing prior work on why some argument is wrong or something
alignment 101 faq on lesswrong for the most common idiotic twitter examples and their rebuttals
better search (lol)
a skill tree of rationality
So: How does
§1. There should be grayspaces around rationality
§Non-lesswrong
§Lesswrong
§1
I don’t know what the relevant parts of the sequences are to understand things critical mass to
TL;DR: I think we as rationalists are doing an okay job at keeping the current community happy (and are even growing, not like most of the other places out there on the internet), but we’re not winning.
Let’s explore what we’d need to change that.
30 days ago, I’ve arrived in SFO.
Ever since I’ve read The Sequences a bit more than two years ago now, I’ve been recommending parts of them ever since—and that has only increased over time.
Over the last few years, I’ve thought a lot about
§Appendix
§The skills we want to teach
§0. What are the skills of a rationalist?
(basically everything from CFAR, and the virtues, and so on. But here are some we could focus on)
- dissecting an argument, being able to call out the wrong parts/missing connections (evenness)
- noticing confusion/a flaw in their thinking, and then stopping (lightness)
- having a productive fight and then changing their own opinion
- wanting to know the truth, how things actually work
- very intuitively actually looking at how the world is
- being good at correcting/noticing your own errors
All of that, taken together, we call ‘seeking the truth’ and ‘winning’.
And as far as I know the Instead of hoping they seep through over time, I propose that we teach them.
What could that look like?
Well, my title gave it already away: I think the right shape for this problem is many, many more grayspaces, and all of them proper onboarding platforms to the discussion.
End: