I see basically every post that gets submitted to LessWrong, whereas many users come in and read things more occasionally, so I thought I’d list 10 posts I like in the 2018 Review that people might have missed.
I wasn’t able to write quick reviews of each of them, so this is more like a list of nominations. I’ve left out a few posts that I expect will naturally be very popular.
This points to a variety of specific mechanisms and gears by which culture and aesthetics affect our judgments and our choices, and combines them into an essay that walked me much further than I was before-hand in noticing these effects. I think it might be in my top 5 posts of 2018.
This has some powerful arguments about when not to say things out loud—when not to make background assumptions explicit—which I think is a really powerful datapoint for a rationalist to take on board. And the anti-Nazi ‘Simple Sabotage Field Manual’ is amazing.
These together are maybe the first time I really understood what Paul’s ideas were. Really helpful. I’m mostly talking about the Eliezer-Paul dialogue, both on Eliezer’s post, and in the comments of Alex’s post. I appreciate zhukeepa putting in the explanatory work post as a necessary step for that dialogue to continue.
So. Many. Examples. This is a strong step forward in conceptualising arguments and disagreements, written by someone who’s read and been involved in an incredible amount of good (and bad) ones on the internet.
This post says a lot about how formal communication ignores many messy parts of what it’s actually like being a human, that helps show how what communication needs to be changes at scale.
One of the most important ideas IMO when thinking about AI and AI alignment. Best summarised by the line: “I am not just saying that adversarial optimization makes small probabilities of failure large. I am saying that in general any optimization at all messes with small probabilities and errors drastically.”
This post communicates a lot of Zvi’s taste and experience about how to design basic markets—prediction markets, in particular. Shaped a lot of my thinking about this subject, and has been a good guide when I’ve been (a little bit) involved in trying to build forecasting infrastructure.
I’d be interested to see posts that other users really liked.
10 posts I like in the 2018 Review
I see basically every post that gets submitted to LessWrong, whereas many users come in and read things more occasionally, so I thought I’d list 10 posts I like in the 2018 Review that people might have missed.
I wasn’t able to write quick reviews of each of them, so this is more like a list of nominations. I’ve left out a few posts that I expect will naturally be very popular.
Naming the Nameless by Sarah Constantin
This points to a variety of specific mechanisms and gears by which culture and aesthetics affect our judgments and our choices, and combines them into an essay that walked me much further than I was before-hand in noticing these effects. I think it might be in my top 5 posts of 2018.
Explicit and Implicit Communication by lionhearted
This has some powerful arguments about when not to say things out loud—when not to make background assumptions explicit—which I think is a really powerful datapoint for a rationalist to take on board. And the anti-Nazi ‘Simple Sabotage Field Manual’ is amazing.
Challenges to Christiano’s capability amplification proposal + Paul’s research agenda FAQ by Eliezer Yudkowsky and zhukeepa (respectively)
These together are maybe the first time I really understood what Paul’s ideas were. Really helpful. I’m mostly talking about the Eliezer-Paul dialogue, both on Eliezer’s post, and in the comments of Alex’s post. I appreciate zhukeepa putting in the explanatory work post as a necessary step for that dialogue to continue.
Varieties of Argumentative Experience by Scott Alexander
So. Many. Examples. This is a strong step forward in conceptualising arguments and disagreements, written by someone who’s read and been involved in an incredible amount of good (and bad) ones on the internet.
Unrolling social metacognition: Three levels of meta are not enough. by Academian
This very clearly lays out the iterative process by which social emotions and attitudes are built up. Foundational for a lot of social modelling.
Act of Charity by Jessicata
A thoughtful dialogue throwing into contrast a lot of key intuitions around deception, self-deception, and ethics.
Public Positions and Private Guts by Vaniver
This post says a lot about how formal communication ignores many messy parts of what it’s actually like being a human, that helps show how what communication needs to be changes at scale.
Optimization Amplifies by Scott Garrabrant
One of the most important ideas IMO when thinking about AI and AI alignment. Best summarised by the line: “I am not just saying that adversarial optimization makes small probabilities of failure large. I am saying that in general any optimization at all messes with small probabilities and errors drastically.”
Intelligent Social Web by Valentine
Really well-written explanation of a basic way of understanding social dynamics, that helps me understand a lot of my experiences.
Prediction Markets: When Do They Work by Zvi
This post communicates a lot of Zvi’s taste and experience about how to design basic markets—prediction markets, in particular. Shaped a lot of my thinking about this subject, and has been a good guide when I’ve been (a little bit) involved in trying to build forecasting infrastructure.
I’d be interested to see posts that other users really liked.