Good posts you might want to nominate in the 2018 Review
I’m on track to nominate around 30 posts from 2018, which is a lot. Here is a list of about 30 further posts I looked at that I think were pretty good but didn’t make my top list, in the hopes that others who did get value out of the posts will nominate their favourites. Each post has a note I wrote down for myself about the post.
This felt like a clear analysis of an idea and coming up with some hypotheses. I don’t think the hypotheses really captures what’s going on, and most of the frames here seem like they’ve caused a lot of people to do a lot of hurt to themselves, but it seemed like progress in that conversation.
This seems to me like close to an important point but not quite saying it. I don’t know if I got anything especially knew from its framing, but its examples are pretty good.
I think that the OP is an important idea. I think my comment on it is pretty good (and the discussion below it), though I’ve substantially changed my position since then, and should write up my new worldview once my life calms down. I don’t think I should nominate it because I’m a major part of the discussion.
Why science has made a taboo of causality feels like a really important question to answer when figuring out how much to trust academia and how to make institutions that successfully make scientific progress, and this post suggests some interesting hypotheses.
A bunch of novel hypotheses about consciousness in different animals that I’d never heard before, which seem really useful for thinking about the topic.
I thought this was a really impressive post, going around and building simple models of lots of different theories, and giving a bit of personal experience with the practitioners of the theories. It was systematic and goal oriented and clear.
There’s a deep problem here of figuring out norms in novel and weird and ambiguous environments in the modern world, especially given the internet, and this post is kind of like a detailed, empirical study of some standard clusters of norms, which I think is very helpful.
Feels like a clear example of a larger and important strategy about changing the incentives on you. Not clear how valuable the pots is alone, but I like it a lot.
I feel like this does a lot of good intuition building work, and I think about this post from time to time in my own life. I think that Jan brought up some good points in the comments about not wanting to cause confusion about different technical concepts all being the same, so I’d like to see the examples reviewed to check they’re all about attachment effects and not conflating different effects.
This makes a really important point anyone learns in the study of mathematics, and I think is generally an important distinction to have understand between language and reality. Just because we have words for some things doesn’t make them more real than things we don’t have words for. The point is to look at reality, not to look at the words.
I think back to this occasionally. Seems like quite a useful distinction, and maybe we should try to encourage people making more guidelines. Maybe we should build a wiki and have a page type ‘guideline’ where people contribute to make great guidelines.
This is one of the first posts that impressed upon me the deeply tangled difficulties of information security, something I’ve increasingly thought a lot about in the intervening years, and expect to think about even more in the future.
Some important conceptual work fundamental to mathematics. Very short and insightful. Not sure if I should allow this though, because if I do am I just allowing all high-level discussion of math to be included?
It feels like useful advice and potentially a valuable observation with which to view a deeper problem. But unclear on the last one, and not sure if this post should be nominated just on the first alone.
The most interesting part about this is the claim that most people who housed Jews during the holocaust did it because the social situation made the moral decision very explicit and that they felt they only had one possible outcome, not because they were proactive moral planners. I would like to see an independent review of this info.
Seems helpful but weird to nominate, as the book is short and this post explicitly doesn’t contain all the key ideas in the book. I did learn from this that having lots of nukes is more stable than having a small number, and this has stuck with me.
I felt great about it when I read this post last time. I’ve not given it a careful re-read, would like to see it reviewed, but I think it’s likely I’ll rediscover it’s a very helpful abstraction.
AI alignment posts you might want to nominate
[Edit: On reflection, I think that the Alignment posts that do not also have implications for human rationality aren’t important to go through this review process, and we’ll likely create another way to review that stuff and make it into books.]
There was also a lot of top-notch AI alignment writing, but I mostly don’t feel well-placed to nominate it. I hope others can look through and nominate selections from these.
Feels like one of the few practical posts that can help a large number of people do embedded agency research, so really valuable from that perspective.
Good posts you might want to nominate in the 2018 Review
I’m on track to nominate around 30 posts from 2018, which is a lot. Here is a list of about 30 further posts I looked at that I think were pretty good but didn’t make my top list, in the hopes that others who did get value out of the posts will nominate their favourites. Each post has a note I wrote down for myself about the post.
Reasons compute may not drive AI capabilities growth
I don’t know if it’s good, but I’d like it to be reviewed to find out.
The Principled-Intelligence Hypothesis
Very interesting hypothesis generation. Unless it’s clearly falsified, I’d like to see it get built on.
Will AI See Sudden Progress? DONE
I think this post should be considered paired with Paul’s almost-identical post. It’s all exactly one conversation.
Personal Relationships with Goodness
This felt like a clear analysis of an idea and coming up with some hypotheses. I don’t think the hypotheses really captures what’s going on, and most of the frames here seem like they’ve caused a lot of people to do a lot of hurt to themselves, but it seemed like progress in that conversation.
Are ethical asymmetries from property rights?
Again, another very interesting hypothesis.
Incorrect Hypotheses Point to Correct Observations ONE NOMINATION
This seems to me like close to an important point but not quite saying it. I don’t know if I got anything especially knew from its framing, but its examples are pretty good.
Whose reasoning can you rely on when your own is faulty?
I really like the questions, and should ask them more about the people I know.
Inconvenience is Qualitatively Bad ONE NOMINATION
I think that the OP is an important idea. I think my comment on it is pretty good (and the discussion below it), though I’ve substantially changed my position since then, and should write up my new worldview once my life calms down. I don’t think I should nominate it because I’m a major part of the discussion.
The abruptness of nuclear weapons
Clearly valuable historical case, simple effect model.
Book Review: Pearl’s Book of Why
Why science has made a taboo of causality feels like a really important question to answer when figuring out how much to trust academia and how to make institutions that successfully make scientific progress, and this post suggests some interesting hypotheses.
Functional Institutions Are the Exception ONE NOMINATION
Was a long meditation on an important idea, that I’ve found valuable to read. Agree with commenter that it’s sorely lacking in examples however.
Strategies of Personal Growth ONE NOMINATION
Oli curated it, he should consider nominating and saying what he found useful. It all seemed good but I didn’t personally get much from it.
Preliminary Thoughts on Moral Weight DONE
A bunch of novel hypotheses about consciousness in different animals that I’d never heard before, which seem really useful for thinking about the topic.
Theories of Pain
I thought this was a really impressive post, going around and building simple models of lots of different theories, and giving a bit of personal experience with the practitioners of the theories. It was systematic and goal oriented and clear.
Clarifying “AI Alignment” DONE
Rohin’s comment is the best part of this post, not sure how best to nominate it.
Norms of Membership for Voluntary Groups
There’s a deep problem here of figuring out norms in novel and weird and ambiguous environments in the modern world, especially given the internet, and this post is kind of like a detailed, empirical study of some standard clusters of norms, which I think is very helpful.
How Old is Smallpox?
Central example of “Things we learned on LessWrong in 2018”. Should be revised though.
“Cheat to Win”: Engineering Positive Social Feedback
Feels like a clear example of a larger and important strategy about changing the incentives on you. Not clear how valuable the pots is alone, but I like it a lot.
Track-Back Meditation
I don’t know why, but I think about this post a lot.
Meditations on Momentum
I feel like this does a lot of good intuition building work, and I think about this post from time to time in my own life. I think that Jan brought up some good points in the comments about not wanting to cause confusion about different technical concepts all being the same, so I’d like to see the examples reviewed to check they’re all about attachment effects and not conflating different effects.
On Exact Mathematical Formulae
This makes a really important point anyone learns in the study of mathematics, and I think is generally an important distinction to have understand between language and reality. Just because we have words for some things doesn’t make them more real than things we don’t have words for. The point is to look at reality, not to look at the words.
Recommendations vs Guidelines
I think back to this occasionally. Seems like quite a useful distinction, and maybe we should try to encourage people making more guidelines. Maybe we should build a wiki and have a page type ‘guideline’ where people contribute to make great guidelines.
On the Chatham House Rule DONE
This is one of the first posts that impressed upon me the deeply tangled difficulties of information security, something I’ve increasingly thought a lot about in the intervening years, and expect to think about even more in the future.
Different types (not sizes!) of infinity
Some important conceptual work fundamental to mathematics. Very short and insightful. Not sure if I should allow this though, because if I do am I just allowing all high-level discussion of math to be included?
Expected Pain Parameters
It feels like useful advice and potentially a valuable observation with which to view a deeper problem. But unclear on the last one, and not sure if this post should be nominated just on the first alone.
Research: Rescuers During the Holocaust DONE
The most interesting part about this is the claim that most people who housed Jews during the holocaust did it because the social situation made the moral decision very explicit and that they felt they only had one possible outcome, not because they were proactive moral planners. I would like to see an independent review of this info.
Lessons from the cold war on information hazards: why internal communication is critical DONE
Seems like an important historical lesson.
Problem Solving With Crayons and Mazes
I didn’t find it much useful. Oli was excited when he curated it, should poke him to consider nominating it.
Insights from “Strategy of Conflict”
Seems helpful but weird to nominate, as the book is short and this post explicitly doesn’t contain all the key ideas in the book. I did learn from this that having lots of nukes is more stable than having a small number, and this has stuck with me.
The Bat and Ball Problem Revisited DONE
A curiosity-driven walk through what’s going on with the bat-and-the-ball problem by Kahneman.
Good Samaritans in Experiments
A highly opinionated and very engaging criticism of a study.
Hammertime Final Exam ONE NOMINATION
Was great, I’m not actually sure whether it fits into this review process?
Naming the Nameless DONE
Some people seemed to get a lot out of this, but I haven’t had the time to engage with it much.
Actually, just re-read it, and it’s brilliant, and one of the best 5-10 of the year. Will nominate it myself if nobody else does.
How did academia ensure papers were correct in the early 20th century? DONE
I’m glad I put this down in writing. I found it useful myself. But others should figure out whether to nominate.
Competitive Markets as Distributed Backdrop DONE
I felt great about it when I read this post last time. I’ve not given it a careful re-read, would like to see it reviewed, but I think it’s likely I’ll rediscover it’s a very helpful abstraction.
AI alignment posts you might want to nominate
[Edit: On reflection, I think that the Alignment posts that do not also have implications for human rationality aren’t important to go through this review process, and we’ll likely create another way to review that stuff and make it into books.]
There was also a lot of top-notch AI alignment writing, but I mostly don’t feel well-placed to nominate it. I hope others can look through and nominate selections from these.
Rohin’s sequence
Paul’s sequence
Abram’s writing that year
Wei Dai’s writing that year
Scott’s “Fixed Point Discussion”
Feels like one of the few practical posts that can help a large number of people do embedded agency research, so really valuable from that perspective.
Vika’s “Discussion on the Machine Learning Approach to AI safety”