I’m an admin of this site; I work full-time on trying to help people on LessWrong refine the art of human rationality.
Longer bio: www.lesswrong.com/posts/aG74jJkiPccqdkK3c/the-lesswrong-team-page-under-construction#Ben_Pace___Benito
Ben Pace(Ben)
Karma: 18,226
I read it as “Messr” as in
“Messrs Moony, Wormtail, Padfoot, and Prongs
Purveyors of Aids to Magical Mischief-Makers
are proud to present
THE MARAUDER’S MAP”
I vouch that this person is both a LW user who has written IMO some good posts and a member of in-person rationalist/longtermist/EA communities who is in good standing.
That sounds like a good response to the Is-Ought problem.
over a dozen other testimonies from former residents and apprentices that have left the monastery over the last 2-3 years in traumatic states
Have you personally talked to over 12 people who have left the monastic academy in the past 2-3 years and they each reported their experience was strongly negative or overall negative?
It is as if we were stuck in a kind of awkward middle ground
Throughout this post I was thinking about the phrase “starting from the middle” as where we are, with neither perfect knowledge of reasoning nor of our goals, but having something to go on and having to build from there.
It also reminds me of Abram’s post An Orthodox Case Against Utility Functions which is about getting started from the point of view of an agent, created in motion, not the point of view of the universe (or some other omniscient sort of position).
This is true inasmuch as posts written with help from GPT-3 are meant to be evidence about the capabilities of GPT-3.
Sometimes posts are primarily intended to be fun, and success is measured by how fun they are, and then I don’t care how much iteration you put into it, I just want it to be fun.
I guess this was a combo, because it’s about simulation? So your question is reasonable. FYI from having played with GPT-3 myself, I assumed something like Lsusr had run multipl (3-15) iterations and partial iterations on each segment and thrown bits out and thrown whole other segments out. That said it was probably clearer to me because I’ve written with GPT-3 myself, and someone who hasn’t could’ve been under the impression this was just the first pass.
Thanks very much, I really liked reading this essay. I concur with your arguments about why more optionality and privacy don’t solve the problem. I also came up with the idea of more competition. That sounds like the sort of solution the market is good at, but I can’t think of a schema for getting around the network effects that you talk about.
(Actually, I just thought of one. I think that if the recommender systems were open so that anyone could write an algorithm, that could lead to pretty good competition. I’d be excited to see alt FB or Twitter algorithms. That said it sounds like more optionality I.e. most people wouldn’t use it. So not sure.)
The auditing is an idea I hadn’t heard before. It reminds me of what Zvi Mowshowitz did for me when he “audited” the FB algorithm ( https://www.google.com/amp/s/thezvi.wordpress.com/2017/04/22/against-facebook/amp/ ). That was very helpful. I’d love to see more work like that.
(This is a pretty interesting incentives angle I hadn’t heard before.)
how is this so good!
Oh yeah, +1 to this. I was surprised the post only had 11 karma when I saw it (William had sent me an advance copy and I’d really liked reading it) but when I saw that it was a link post I understood why.
There’s a related dynamic that came up in a convo I just had.
Alice: My current work is exploring if we can solve value loading using reward learning.
Bob: Woah, isn’t that obviously doomed? Didn’t Rohin write a whole sequence on this?
Alice: Well, I don’t want to solve the whole problem for arbitrary difficulty. I just want to know whether we can build something that gets the basics right in distributions that a present day human can understand. For example I reckon we may be able to teach an AI what murder is today, even if we can’t teach it what murder is post-singularity.
Bob: I see. That’s more reasonable. However, there is a school of thought that I suspect you may be a part of, called “slow takeoff”, that looks like over the course of maybe 10 years the world will increasingly be reliant on ML systems whose internals we don’t understand, and whose actions we don’t understand (we just see the metrics go up). It is world is already out of distribution for what a present day human can understand.
Alice: That’s true. I guess I’m interested to know how far we can push it and still build a system that helps us take pivotal acts and doesn’t destroy everything.
(The above is my paraphrase. Both said different things than the above and also I incorporated some things I said.)
The success story you have in mind determines what problem you’re trying to solve, and to some extent which capability regime you’re thinking about.
From the OP it doesn’t seem to me like the author is saying that. It’s not all there in that quote, but put it in the context of this part:
when DiAngelo sees white people at such meetings clamming up, or talking in meandering ways, with “long pauses” and “self-corrections”, she just chalks this up as white people deploying their favorite strategies for perpetuating white privilege.
I suspect the rule here being taught is: do not reflect, do not think, do not qualify. These are all ways that you are politically opposing us. As before, “the best response is to say “of course I was; I’m glad I finally found out about it so I can change.””
Naturally, it’s not v LW to tell people not to think in conversations or correct previous inaccurate statements, and to tell them instead what to say and that the most acceptable outcome is to just agree on the object level.
I strong upvoted it initially when I assumed the state of affairs was “nobody had done the work to make a recording+transcript, so this post is helping out”. Then I switched to strong downvote when I read Bjartur’s comment saying Sam had requested no transcript or recording.
Thanks for making the change!
When you find out you’ve been doing something that perpetuates racismS, the best response is to say “of course I was; I’m glad I finally found out about it so I can change.”
It seems to me this book is largely a manual for obedience to a political faction; a long list of the details of how one ought to act in different scenarios in order to signal obeisance.
Edit: I previously started with the the line “To comply, I have no choice but to immediately agree on the object level and submit.” But this is an exaggeration of what is written in the quote, so I’m removing it / adding this edit note.
I have more than 100 karma and still get a lot of value out of feedback on posts.
Hm, seems like an obvious disincentive for speaking at ACX events if, on requesting no recording+transcript, someone writes up notes of what you said and publishes them on the web.
p.b., I propose that you change the title to “Rough Notes: Sam Altman Q&A on GPT + AGI”. I think this will help people not be confused about how much the notes are claiming to be accurate – it seems the notes are ‘roughly’ accurate but with some mistakes. As Sam’s a prominent person I think you can expect small mistakes to flow out far and it’s good to avoid that.
This comment.
What’s the value you get from it, and how does this once-a-year event affect the value you get from LW?