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.)
I generally feel more hopeful about a situation when I understand it better.
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.)
I generally feel more hopeful about a situation when I understand it better.
I don’t think it applies to safety researchers at AI Labs though, I am shocked how much those folks can make.
A common experience I have is that it takes like 1-2 paragraphs of explanation for why I want this info (e.g. “Well I’m wondering if so-and-so should fly in a day earlier to travel with me but it requires going to a different airport and I’m trying to figure out whether the time it’d take to drive to me would add up to too much and also...”), but if they just gave me their ~70% confidence interval when I asked then we could cut the whole context-sharing.
Often I am annoyed when I ask someone (who I believe has more information than me) a question and they say “I don’t know”. I’m annoyed because I want them to give me some information. Such as:
“How long does it take to drive to the conference venue?”
“I don’t know.”
“But is it more like 10 minutes or more like 2 hours?”
“Oh it’s definitely longer than 2 hours.”
But perhaps I am the one making a mistake. For instance, the question “How many countries are there?” can be answered “I’d say between 150 and 400″ or it can be answered “195”, and the former is called “an estimate” and the latter is called “knowing the answer”. There is a folk distinction here and perhaps it is reasonable for people to want to preserve the distinction between “an estimate” and “knowing the answer”.
So in the future, to get what I want, I should say “Please can you give me an estimate for how long it takes to drive to the conference venue?”.
And personally I should strive, when people ask me a question to which I don’t know the answer, to say “I don’t know the answer, but I’d estimate between X and Y.”
I think it is pretty obviously a joke :P
(And in case anyone was led astray: the Marx quote at the start is from Groucho, not Karl.)
K. I recommend that people include links for those of us who mostly do not read Twitter.
Crossposted from where?
Curated! Very interesting to get a vivid sense of what goes on when people are facing strong pressures to lie, and how they go about doing this. Both their adamance that they were right and their transparency to you were both fascinating. And this was very engagingly written. Thanks for the post!
As someone who’s spent a while designing charts for published books, I have generally been strongly against axis lines. One thing that has really influenced my approach to using lines is the section of Butterick’s Practical Typography on tables.
Nowadays I remove all lines on tables and charts unless there’s a strong argument in favor of one; implied lines are much easier on the eye.
This post overall moved me toward using gridlines a little bit more, for intuitively measuring distance when that’s important.
I think this essay raises many good points, but doesn’t grapple with (to me) the hardest part of wholesomeness: when do I ignore parts of the whole?
I think that sometimes you make the choice not think about something for a while. For instance, trivially, you can only track so many hypotheses in detail. While I am designing a product that I think will change the world, I will spend most of my time considering different hypotheses for what sort of product users want, and considering how to quickly falsify them and iterate. I will not spend a ton of time questioning whether capitalism is even good for civilization. Insofar as I’m choosing to give this product a shot, that is not a good use of mental resources—the assumption questioning comes before, and after (and occasionally in the middle of I have exceptional cause for a crisis of faith).
To me the hard question of wholesomeness is about knowing when you’re choosing look away from a thing because on reflection it’s not worth the cognitive space to be tracking it as a consideration, and knowing when you’re doing it improperly because it’s painful or emotionally draining or personally inconvenient to keep looking at the thing.
(And that emotional cost itself is a factor to be weighed on the scales.)
Some written guidance on this would be valuable, I’d say.
I assign >10% that Anthropic will at some point pause development for at least a year as a result of safety evaluations.
I think that if Anthropic cannot make a certain product-line safe and then they pivot to scaling up a different kind of model / product-line, I am not counting this as ‘pausing development’.
If they pause all novel capability development and scaling and just let someone else race ahead while pivoting to some other thing like policy or evals or something (while continuing to iterate on existing product lines) then I am counting that as pausing development.
I’ve added it back in. Seemed like a fairly non-specific word to me.
Plausible. I was imitating the phrasing used by an Anthropic funder here. I’m open to editing it in the next hour or so if you think there’s a better phrasing.
I’m interested to know why you think that. I’ve not thought about it a ton so I don’t think I’d be a great dialogue partner, but I’d be willing to give it a try, or you could give an initial bulleted outline of your reasoning here.
I assign >20% that many of the Anthropic employees who quit OpenAI signed Non-Disparagement Agreements with OpenAI.
Anthropic has (in expectation) brought forward the date of superintelligent AGI development (and not slowed it down).
I assign >50% probability to the claim that Anthropic will release products far beyond the frontier within the next 5 years.
I assign >20% probability to the claim that Anthropic will release products far beyond the frontier within the next 5 years.
I’d say most people assume I want “the answer” rather than “some bits of information”.