What are your or Vassar’s arguments against EA or AI alignment? This is only tangential to your point, but I’d like to know about it if EA and AI alignment are not important.
The general argument is that EA’s are not really doing what they say they do. One example from Vassar would be that when it comes to COVID-19 for example there seem to be relatively little effective work by EA’s. In contrast Vassar considered giving prisoners access to personal equipment the most important and organized effectively for that to happen.
EA’s created in EA Global an enviroment where someone who wrote a good paper warning about the risks of gain-of-function research doesn’t address that directly but only talks indirectly about it to focus on more meta-issues. Instead of having conflicts with people doing gain-of-function research the EA community mostly ignored it’s problems and funded work that’s in less conflict with the establishment. There’s nearly no interest in learning from those errors in the EA community and people rather avoid conflicts.
If you read the full comments of this thread you will find reports that CEA used legal threats to cover up Leverage related information.
AI alignment is important but just because one “works on AI risk” doesn’t mean that the work actually decreases AI risk. Tying your personal identity to being someone who works to decrease AI risk makes it hard to clearly reason about whether one’s actions actually do. OpenAI would be an organization where people who see themselves as “working on AI alignment” work and you can look at the recent discussion that whether or not that work reduces or increases actual risk is in open debate.
In a world where human alignment doesn’t work to prevent dangerous gain of function experiments from happening thinking about AI alignment instead of the problem of human alignment where it’s easier to get feedback might be the wrong strategic focus.
Did Vassar argue that existing EA organizations weren’t doing the work they said they were doing, or that EA as such was a bad idea? Or maybe that it was too hard to get organizations to do it?
(a) EA orgs aren’t doing what they say they’re doing (e.g. cost effectiveness estimates are wildly biased, reflecting bad procedures being used internally), and it’s hard to get organizations to do what they say they do
(b) Utilitarianism isn’t a form of ethics, it’s still necessary to have principles, as in deontology or two-level consequentialism
(c) Given how hard it is to predict the effects of your actions on far-away parts of the world (e.g. international charity requiring multiple intermediaries working in a domain that isn’t well-understood), focusing on helping people you have more information about makes sense unless this problem can be solved
(d) It usually makes more sense to focus on ways of helping others that also build capacities, including gathering more information, to increase long-term positive impact
If you for example want the critcism on GiveWell, Ben Hoffman was employed at GiveWell and made experiences that suggest that the process based on which their reports are made has epistemic problems. If you want the details talk to him.
The general model would be that between actual intervention and the top there are a bunch of maze levels. GiveWell then hired normal corporatist people who behave in the dynamics that the immoral maze sequence describes play themselves out.
Vassar’s action themselves are about doing altruistic actions more directly by looking for who are most powerless who need help and working to help them. In the COVID case he identified prisoners and then worked on making PPE available for them.
You might see his thesis is that “effective” in EA is about adding a management layer for directing interventions and that management layer has the problems that the immoral maze sequence describes. According to Vassar someone who wants to be altrustic shouldn’t delegate his judgements of what’s effective and thus warrents support to other people.
I have a large number of negative Leverage experiences between 2015-2017 that I never wrote up due to various complicated adversarial dynamics surrounding Leverage and CEA (as well as various NDAs and legal threats, made by both Leverage and CEA, not leveled at me, but leveled at enough people around me that I thought I might cause someone serious legal trouble if I repeat a thing I heard somewhere in a more public setting)
I’m getting a bit pedantic, but I wouldn’t gloss this as “CEA used legal threats to cover up Leverage related information”. Partly because the original bit is vague, but also because “cover up” implies that the goal is to hide information.
For example, imagine companies A and B sue each other, which ends up with them settling and signing an NDA. Company A might accept an NDA because they want to move on from the suit and agreeing to an NDA does that most effectively. I would not describe this as company A using legal threats to cover up B-related information.
In the timeframe CEA and Leverage where doing together the Pareto Fellowship. If you read the common knowledge post you find people finding that they were mislead by CEA because the announcement didn’t mention that the Pareto Fellowship was largely run by Leverage.
On their mistakes page CEA, they have a section about the Pareto Fellowship but it hides the fact that Leverage was involved in the Pareto Fellowship but says “The Pareto Fellowship was a program sponsored by CEA and run by two CEA staff, designed to deepen the EA involvement of promising students or people early in their careers.”
That does look to me like hiding information about the cooperation between Leverage and CEA.
I do think that publically presuming that people who hide information have something to hide is useful. If there’s nothing to hide I’d love to know what happened back then or who thinks what happened should stay hidden. At the minimum I do think that CEA witholding the information that the people who went to their programs spend their time in what now appears to be a cult is something that CEA should be open about in their mistakes page.
Yep, I think CEA has in the past straightforwardly misrepresented (there is a talk on the history of EA by Will and Toby that says some really dubious things here, IIRC) and sometimes even lied in order to not mention Leverage’s history with Effective Altruism. I think this was bad, and continues to be bad.
My initial thought on reading this was ‘this seems obviously bad’, and I assumed this was done to shield CEA from reputational risk.
Thinking about it more, I could imagine an epistemic state I’d be much more sympathetic to: ‘We suspect Leverage is a dangerous cult, but we don’t have enough shareable evidence to make that case convincingly to others, or we aren’t sufficiently confident ourselves. Crediting Leverage for stuff like the EA Summit (without acknowledging our concerns and criticisms) will sound like an endorsement of Leverage, which might cause others to be drawn into its orbit and suffer harm. But we don’t feel confident enough to feel comfortable tarring Leverage in public, or our evidence was shared in confidence and we can’t say anything we expect others to find convincing. So we’ll have to just steer clear of the topic for now.’
Still seems better to just not address the subject if you don’t want to give a fully accurate account of it. You don’t have to give talks on the history of EA!
I think the epistemic state of CEA was some mixture of something pretty close to what you list here, and something that I would put closer to something more like “Leverage maybe is bad, or maybe isn’t, but in any case it looks bad, and I don’t think I want people to think EA or CEA is bad, so we are going to try to avoid any associations between these entities, which will sometimes require stretching the truth”.
“Leverage maybe is bad, or maybe isn’t, but in any case it looks bad, and I don’t think I want people to think EA or CEA is bad, so we are going to try to avoid any associations between these entities, which will sometimes require stretching the truth”
That has the collary: “We don’t expect EA’s to care enough about the truth/being transparent that this is a huge reputational risk for us.”
It does look weird to me that CEA doesn’t include this on the mistakes page when they talk about Pareto. I just sent CEA an email to ask:
Hi CEA,
On https://www.centreforeffectivealtruism.org/our-mistakes I see “The Pareto Fellowship was a program sponsored by CEA and run by two CEA staff, designed to deepen the EA involvement of promising students or people early in their careers. We realized during and after the program that senior management did not provide enough oversight of the program. For example, reports by some applicants indicate that the interview process was unprofessional and made them deeply uncomfortable.”
Is there a reason that the mistakes page does not mention the involvement of Leverage in the Pareto Fellowship? [1]
Yep, I think the situation is closer to what Jeff describes here, though, I honestly don’t actually know, since people tend to get cagey when the topic comes up.
I talked with Geoff and according to him there’s no legal contract between CEA and Leverage that prevents information sharing. All information sharing is prevented by organization internal NDA’s.
Huh, that’s surprising, if by that he means “no contracts between anyone currently at Leverage and anyone at CEA”. I currently still think it’s the case, though I also don’t see any reason for Geoff to lie here. Maybe there is some technical sense in which there is no contract between Leverage and CEA, but there are contracts between current Leverage employees, who used to work at CEA, and current CEA employees?
What he said is compatible with Ex-CEA people still being bound by the NDA’s they signed they were at CEA. I don’t think anything happened that releases ex-CEA people from NDAs.
The important thing is that CEA is responsible for those NDA and is free to unilaterally lift them if they would have an interest in the free flow of information. In the case of a settlement with contracts between the two organisations CEA couldn’t unilaterally lift the settlement contract.
Public pressure on CEA seems to be necessary to get the information out in the open.
What are your or Vassar’s arguments against EA or AI alignment? This is only tangential to your point, but I’d like to know about it if EA and AI alignment are not important.
The general argument is that EA’s are not really doing what they say they do. One example from Vassar would be that when it comes to COVID-19 for example there seem to be relatively little effective work by EA’s. In contrast Vassar considered giving prisoners access to personal equipment the most important and organized effectively for that to happen.
EA’s created in EA Global an enviroment where someone who wrote a good paper warning about the risks of gain-of-function research doesn’t address that directly but only talks indirectly about it to focus on more meta-issues. Instead of having conflicts with people doing gain-of-function research the EA community mostly ignored it’s problems and funded work that’s in less conflict with the establishment. There’s nearly no interest in learning from those errors in the EA community and people rather avoid conflicts.
If you read the full comments of this thread you will find reports that CEA used legal threats to cover up Leverage related information.
AI alignment is important but just because one “works on AI risk” doesn’t mean that the work actually decreases AI risk. Tying your personal identity to being someone who works to decrease AI risk makes it hard to clearly reason about whether one’s actions actually do. OpenAI would be an organization where people who see themselves as “working on AI alignment” work and you can look at the recent discussion that whether or not that work reduces or increases actual risk is in open debate.
In a world where human alignment doesn’t work to prevent dangerous gain of function experiments from happening thinking about AI alignment instead of the problem of human alignment where it’s easier to get feedback might be the wrong strategic focus.
Did Vassar argue that existing EA organizations weren’t doing the work they said they were doing, or that EA as such was a bad idea? Or maybe that it was too hard to get organizations to do it?
He argued
(a) EA orgs aren’t doing what they say they’re doing (e.g. cost effectiveness estimates are wildly biased, reflecting bad procedures being used internally), and it’s hard to get organizations to do what they say they do
(b) Utilitarianism isn’t a form of ethics, it’s still necessary to have principles, as in deontology or two-level consequentialism
(c) Given how hard it is to predict the effects of your actions on far-away parts of the world (e.g. international charity requiring multiple intermediaries working in a domain that isn’t well-understood), focusing on helping people you have more information about makes sense unless this problem can be solved
(d) It usually makes more sense to focus on ways of helping others that also build capacities, including gathering more information, to increase long-term positive impact
If you for example want the critcism on GiveWell, Ben Hoffman was employed at GiveWell and made experiences that suggest that the process based on which their reports are made has epistemic problems. If you want the details talk to him.
The general model would be that between actual intervention and the top there are a bunch of maze levels. GiveWell then hired normal corporatist people who behave in the dynamics that the immoral maze sequence describes play themselves out.
Vassar’s action themselves are about doing altruistic actions more directly by looking for who are most powerless who need help and working to help them. In the COVID case he identified prisoners and then worked on making PPE available for them.
You might see his thesis is that “effective” in EA is about adding a management layer for directing interventions and that management layer has the problems that the immoral maze sequence describes. According to Vassar someone who wants to be altrustic shouldn’t delegate his judgements of what’s effective and thus warrents support to other people.
Link? I’m not finding it
https://www.lesswrong.com/posts/MnFqyPLqbiKL8nSR7/my-experience-at-and-around-miri-and-cfar-inspired-by-zoe?commentId=zqcynfzfKma6QKMK9
I think what you’re pointing to is:
I’m getting a bit pedantic, but I wouldn’t gloss this as “CEA used legal threats to cover up Leverage related information”. Partly because the original bit is vague, but also because “cover up” implies that the goal is to hide information.
For example, imagine companies A and B sue each other, which ends up with them settling and signing an NDA. Company A might accept an NDA because they want to move on from the suit and agreeing to an NDA does that most effectively. I would not describe this as company A using legal threats to cover up B-related information.
In the timeframe CEA and Leverage where doing together the Pareto Fellowship. If you read the common knowledge post you find people finding that they were mislead by CEA because the announcement didn’t mention that the Pareto Fellowship was largely run by Leverage.
On their mistakes page CEA, they have a section about the Pareto Fellowship but it hides the fact that Leverage was involved in the Pareto Fellowship but says “The Pareto Fellowship was a program sponsored by CEA and run by two CEA staff, designed to deepen the EA involvement of promising students or people early in their careers.”
That does look to me like hiding information about the cooperation between Leverage and CEA.
I do think that publically presuming that people who hide information have something to hide is useful. If there’s nothing to hide I’d love to know what happened back then or who thinks what happened should stay hidden. At the minimum I do think that CEA witholding the information that the people who went to their programs spend their time in what now appears to be a cult is something that CEA should be open about in their mistakes page.
Yep, I think CEA has in the past straightforwardly misrepresented (there is a talk on the history of EA by Will and Toby that says some really dubious things here, IIRC) and sometimes even lied in order to not mention Leverage’s history with Effective Altruism. I think this was bad, and continues to be bad.
My initial thought on reading this was ‘this seems obviously bad’, and I assumed this was done to shield CEA from reputational risk.
Thinking about it more, I could imagine an epistemic state I’d be much more sympathetic to: ‘We suspect Leverage is a dangerous cult, but we don’t have enough shareable evidence to make that case convincingly to others, or we aren’t sufficiently confident ourselves. Crediting Leverage for stuff like the EA Summit (without acknowledging our concerns and criticisms) will sound like an endorsement of Leverage, which might cause others to be drawn into its orbit and suffer harm. But we don’t feel confident enough to feel comfortable tarring Leverage in public, or our evidence was shared in confidence and we can’t say anything we expect others to find convincing. So we’ll have to just steer clear of the topic for now.’
Still seems better to just not address the subject if you don’t want to give a fully accurate account of it. You don’t have to give talks on the history of EA!
I think the epistemic state of CEA was some mixture of something pretty close to what you list here, and something that I would put closer to something more like “Leverage maybe is bad, or maybe isn’t, but in any case it looks bad, and I don’t think I want people to think EA or CEA is bad, so we are going to try to avoid any associations between these entities, which will sometimes require stretching the truth”.
That has the collary: “We don’t expect EA’s to care enough about the truth/being transparent that this is a huge reputational risk for us.”
It does look weird to me that CEA doesn’t include this on the mistakes page when they talk about Pareto. I just sent CEA an email to ask:
They wrote back, linking me to https://www.lesswrong.com/posts/Kz9zMgWB5C27Pmdkh/common-knowledge-about-leverage-research-1-0?commentId=2QcdhTjqGcSc99sNN
(“we’re working on a couple of updates to the mistakes page, including about this”)
Yep, I think the situation is closer to what Jeff describes here, though, I honestly don’t actually know, since people tend to get cagey when the topic comes up.
I talked with Geoff and according to him there’s no legal contract between CEA and Leverage that prevents information sharing. All information sharing is prevented by organization internal NDA’s.
Huh, that’s surprising, if by that he means “no contracts between anyone currently at Leverage and anyone at CEA”. I currently still think it’s the case, though I also don’t see any reason for Geoff to lie here. Maybe there is some technical sense in which there is no contract between Leverage and CEA, but there are contracts between current Leverage employees, who used to work at CEA, and current CEA employees?
What he said is compatible with Ex-CEA people still being bound by the NDA’s they signed they were at CEA. I don’t think anything happened that releases ex-CEA people from NDAs.
The important thing is that CEA is responsible for those NDA and is free to unilaterally lift them if they would have an interest in the free flow of information. In the case of a settlement with contracts between the two organisations CEA couldn’t unilaterally lift the settlement contract.
Public pressure on CEA seems to be necessary to get the information out in the open.