@Kat Woods
I'm trying to piece together a timeline of events.
You say in the evidence doc that
3 days after starting at Nonlinear, Alice left to spend a whole month with her family. We even paid her for 3 of the 4 weeks despite her not doing much work. (To be fair, she was sick.)
Can you tell me what month this was? Does this mean just after she quit her previous job or just after she started traveling with you?
Why might humans evolve a rejection of things that taste to sweet? What fitness reducing thing does "eating oversweet things" correlate with? Or is it a spandrel of something else?
If this is true, it's fascinating, because it suggest that our preference for cold and carbonation are a kind of specification gaming!
Ok. Given all that, is there particular thing that you wish Ben (or someone) had done differently here? Or are you mostly wanting point out the dynamic?
I want to try to paraphrase what I hear you saying in this comment thread, Holly. Please feel free to correct any mistakes or misframings in my paraphrase.
I hear you saying...
Crostposted from LessWong (link)
Maybe I'm missing something, but it seems like it should take less than an hour to read the post, make a note of every claim that's not true, and then post that list of false claims, even if it would take many days to collect all the evidence that shows those points are false.
I imagine that would be helpful for you, because readers are much more likely to reserve judgement if you listed which specific things are false.
Personally, I could look over that list and say "oh yeah, number 8 [or whatever] is cruxy for me. If t...
(2) I think something odd about the comments claiming that this post is full of misinformation, is that they don't correct any of the misinformation. Like, I get that assembling receipts, evidence etc can take a while, and writing a full rebuttal of this would take a while. But if there are false claims in the post, pick one and say why it's false.
Seconding this.
I would be pretty interested to read a comment from nonlinear folks listing out everything that they believe to be false in the narrative as stated, even if they can't substantiate their counter-claims yet.
I agree that if it were just a few disputed claims that would be a a reasonable thing to do, there are so many. And there is so much nuance.
Here is one example, however. This took us hours to prepare, just to rebut a single false claim:
https://forum.effectivealtruism.org/posts/5pksH3SbQzaniX96b/a-quick-update-from-nonlinear
I recommend that you use a spoiler tag for that last part. Not everyone who wants to has finished the story!
I imagine that most of the disagreement is with (implied, but not stated) conditional "that Owen did this means that decent men don't exist".
I want to know if you can find more people companies that have experienced a similar thing with the FDA.
Is there a reddit or discussion forum where people discuss and commiserate about FDA threats like this one? Can you find people there, and then verify that they / their experiences are real?
As a naive outsider, it seems to me like all of the specific actions you suggest would be stronger and more compelling if you can muster a legitimate claim that this is a pattern of behavior and not just a one-off. An article with one source making an accusation...
I know that I was wrong because people of the global majority continuously speak in safe spaces about they feel unsafe in EA spaces. They speak about how they feel harmed by the kinds of things discussed in EA spaces. And they speak about how there are some people — not everyone, but some people — who don’t seem to be just participating in open debate in order to get at the truth, but rather seem to be using the ideals of open discussion to be a cloak that can hide their intent to do harm.
I'm not sure what to say to this.
A...
Some speech is harmful. Even speech that seems relatively harmless to you might be horribly upsetting for others. I know this firsthand because I’ve seen it myself.
I want to distinguish between "harmful" and "upsetting". It seems to me that there is a big difference between shouting 'FIRE' in a crowed theater, "commanding others to do direct harm" on the one hand, and "being unable to focus for hours" after reading a facebook thread, being exhausted from fielding questions.
My intuitive grasp of the...
First of all, I took this comment to be sincere and in the spirit of dialog. Thank you and salutations.
[Everything that I say in this comment is tentative, and I may change my mind.]
Surely there exists a line at which we agree on principle. Imagine that, for example, our EA spaces were littered with people making cogent arguments that steel manned holocaust denial, and we were approached by a group of Jewish people saying “We want to become effective altruists because we believe in the stated ideals, but we don’t feel safe participating...
I don't follow how what you're saying is a response to what I was saying.
I think a model by which people gradually "warm up" to "more advanced" discourse norms is false.
I wasn't saying "the point of different discourse norms in different EA spaces is that it will gradually train people into more advanced discourse norms." I was saying if that I was mistaken about that "warming up effect", it would cause me to reconsider my view here.
In the comment above, I am only saying that I think it is a mistake to have different discourse norms at the core vs. the periphery of the movement.
I think there is a lot of detail and complexity here and I don't think that this comment is going to do it justice, but I want to signal that I'm open to dialog about these things.
For example, allowing introductory EA spaces like the EA Facebook group or local public EA group meetups to disallow certain forms of divisive speech, while continuing to encourage serious open discussion in more advanced EA spaces, like on this EA forum.
On the face of it, this seems like a bad idea to me. I don't want "introductory" EA spaces to have ...
Surely there exists a line at which we agree on principle. Imagine that, for example, our EA spaces were littered with people making cogent arguments that steel manned holocaust denial, and we were approached by a group of Jewish people saying “We want to become effective altruists because we believe in the stated ideals, but we don’t feel safe participating in a space where so many people commonly and openly argue that the holocaust did not happen.”
In this scenario, I hope that we’d both agree that it would be appropriate for u...
"I think a model by which people gradually "warm up" to "more advanced" discourse norms is false."
I don't think that's the main benefit of disallowing certain forms of speech at certain events. I'd imagine it'd be to avoid making EA events attractive and easily accessible for, say, white supremacists. I'd like to make it pretty costly for a white supremacist to be able to share their ideas at an EA event.
On the most crucial topics, and in capturing the nuance and complexity of the real world, this piece fails again and again: epistemic overconfidence plus uncharitable disdain for the work of others, spread thinly over as many topics as possible.
Interestingly, this reminds me of Nassim Nicholas Taleb.
Another thing for people to keep in mind:
Apparently, if you want loan forgiveness, you can only spend 8 weeks worth of the money on payroll.
From here,
If you’re a sole proprietor, you can have eight weeks of the loan forgiven as a replacement for lost profit. But you’ll need to provide documentation for the remaining two weeks worth of cash flow, proving you spent it on mortgage interest, rent, lease, and utility payments.
So if at some point you need to check boxes saying what you're applying for this loan for, and you can check more ...
I recommend that everyone who is eligible apply through US Bank ASAP.
Other lenders might still work, but US Bank was by far the fastest. A person that I was coaching through this process and I both received our loans within 4 days of initially filling out their application (I say "initially" because there were several steps where they needed additional info).
Also, we now know that the correct answer to how many employees you have is "0 employees, it's just me", not "1 employee, because I employ myself."
An email I received from Bench reads: "If your bank isn’t participating, your next best option is to apply through Fundera—they will match you with the best lender."
However, when I tried to fill out their application, they asked me to upload...
...of which I have only one out of three.
I think a lot of this is right and important, but I especially love:
Don't let the fact that Bill Gates saved a million lives keep you from saving one.
We're all doing the best we can with the privileges we were blessed with.
I like the breakdown of those two bullet points, a lot, and I want to think more about them.
Both of these I think are fairly easily measurable from looking at someone's past work and talking to them, though.
I bet that you could do that, yes. But that seems like a different question than making a scalable system that can do it.
In any case, Ben articulates the view that generated the comment above, above.
[Edit: it'd be very strange if we end up preferring candidates who hadn't thought about AI at all to candidates who had thought some about AI but don't have specific plans for it.]
That doesn't seem that strange to me. It seems to mostly be a matter of timing.
Yes, eventually we'll be in an endgame where the great powers are making substantial choices about how powerful AI systems will be deployed. And at that point I want the relevant decision makers to have sophisticated views about AI risk and astronomical stakes.
But in the the...
At the moment, not really.
There's the classic Double Crux post. Also, here's a post I wrote, that touches on one sub-skill (out of something like 50 to 70 sub-skills that I currently know). Maybe it helps give the flavor.
If I were to say what I'm trying to do in a sentence: "Help the participants actually understand eachother." Most people generally underestimate how hard this is, which is a large part of the problem.
The good thing that I'm aiming for in a conversation is when "that absurd / confused thing that X-person...
I really admire the level of detail and transparency going into these descriptions, especially those written by Oliver Habryka
Hear, hear.
I feel proud of the commitment to epistemic integrity that I see here.
[Are there ways to delete a comment? I started to write a comment here, and then added a bit to the top-level instead. Now I can't make this comment go away?]
A small correction:
Facilitating conversations between top people in AI alignment (I’ve in particular heard very good things about the 3-day conversation between Eric Drexler and Scott Garrabrant that Eli facilitated)
I do indeed facilitate conversations between high level people in AI alignment. I have a standing offer to help with difficult conversations / intractable disagreements, between people working on x-risk or other EA causes.
(I'm aiming to develop methods for resolving the most intractable disagreements in the space. The more direct experien...
(Eli's personal notes, mostly for his own understanding. Feel free to respond if you want.)
1. It seems pretty likely that early advanced AI systems won't be understandable in terms of HRAD's formalisms, in which case HRAD won't be useful as a description of how these systems should reason and make decisions.
My current guess is that the finalized HRAD formalisms would be general enough that they will provide meaningful insight into early advanced AI systems (even supposing that the development of those early systems is not influenced by HRAD ideas), in much the same way that Pearlean causality and Bayes nets gives (a little) insight into what neural nets are doing.
I'm not sure I follow. The question asks what the participants think is most important, which may or may not be diversity of perspectives. At least some people think that diversity of perspectives is a misguided goal, that erodes core values.
Are you saying that this implies that "EA wants more of the same" because some new EA (call him Alex) will be paired with a partner (Barbra) who gives one of the above answers, and then Alex will presume that what Barbra said was the "party line" or "the EA answer" or "what everyone thinks"?
I like these modified questions.
The reason why the original formulations are what they are is to get out of the trap of everyone agreeing that "good things are good", and to draw out specific disagreements.
The intention is that each of these has some sort of crisp "yes or no" or "we should or shouldn't prioritize X". But also the crisp "yes or no" is rooted in a detailed, and potentially original, model.
What sort of discussions does this question generate?
Here are demographics that I've heard people list.
All of these have very different implications about what is most important on the margin in EA.
I think using EA examples in the double crux game may be a bad idea because it will inadvertently lead EAs to come away with a more simplistic impression of these issues than they should.
I mostly teach Double Crux and related at CFAR workshops (the mainline, and speciality / alumni workshops). I've taught it at EAG 4 times (twice in 2017), and I can only observe a few participants in a session. So my n is small, and I'm very unsure.
But it seems to me that using EA examples mostly has the effect of fleshing out understanding of other EA's views, more t...
I strongly agree that more EAs doing independent thinking really important, and I'm very interested in interventions that push in that direction. In my capacity as a CFAR instructor and curriculum developer, figuring out ways to do this is close to my main goal.
I think many individual EAs should be challenged to generate less confused models on these topics, and from there between models is when deliberation like double crux should start.
Strongly agree.
...I don't think in the span of only a couple minutes either side of a double crux game will generate
I'm not sure how much having a "watered down" version of EA ideas in the zeitgeist helps because, I don't have a clear sense of how effective most charities are.
If the difference between the median charity and the most impactful charity is 4 orders of magnitude ($1 to the most impactful charities does as much good as $1000 to the the median charity), then even a 100x improvement from the median charity is not very impactful. It's still only 1% as good a donating to the best charity. If that were the case, it's probably more efficient to just aim...
It seems to me that in many cases the specific skills that are needed are both extremely rare and not well captured by the standard categories.
For instance, Paul Christiano seems to me to be an enormous asset to solving the core problems of AI safety. If "we didn't have a Paul" I would be willing to trade huge amounts of EA resources to have him working on AI safety, and I would similarly trade huge resources to get another Paul-equivalent working on the problem.
But it doesn't seem like Paul's skillset is one that I can easily select for. He's kn...
There aren't many people with PhD-level research experience in relevant fields who are focusing on AI safety, so I think it's a bit early to conclude these skills are "extremely rare" amongst qualified individuals.
AI safety research spans a broad range of areas, but for the more ML-oriented research the skills are, unsurprisingly, not that different from other fields of ML research. There are two main differences I've noticed:
In the short term, senior hires are most likely to come from finding and onboarding people who already have the required skills, experience, credentials and intrinsic motivation to reduce x-risks.
Can you be more specific about the the required skills and experience are?
Skimming the report, you say "All senior hires require exceptionally good judgement and decision-making." Can you be more specific about what that means and how it can be assessed?
It seems to me that in many cases the specific skills that are needed are both extremely rare and not well captured by the standard categories.
For instance, Paul Christiano seems to me to be an enormous asset to solving the core problems of AI safety. If "we didn't have a Paul" I would be willing to trade huge amounts of EA resources to have him working on AI safety, and I would similarly trade huge resources to get another Paul-equivalent working on the problem.
But it doesn't seem like Paul's skillset is one that I can easily select for. He's kn...
Intellectual contributions to the rationality community: including CFAR’s class on goal factoring
Just a note. I think this might be a bit missleading. Geoff, and other members of Leverage research taught a version of goal factoring at some early CFAR workshops. And Leverage did develop a version of goal factoring inspired by CT. But my understanding is that CFAR staff independently developed goal factoring (starting from an attempt to teach applied consequentialism), and this is an instance of parallel development.
[I work for CFAR, though I had not yet joined the EA or rationality community in those early days. I am reporting what other longstanding CFAR staff told me.]
Great. Thank you!