Thanks so much Sofia for building on my post! So good to see a movement leader weighing in. Great observations. To build on yours in turn:
- Love the idea of "context-absorption infrastructure". Interestingly this is something I recommend that organisations do for the opposite reason as well. I think that organisations should be trying to document as much of their context as possible because giving that context to AI allows AI to do better work. I like that this could also have the side benefit of creating a very good set of learning materials to allow new hires to rapidly come up to speed.
- I also think it's an excellent point that the movement might collapse to a smaller number of very experienced people. Something that AI can't really replicate is the tacit experience that comes from having done something every day. That kind of thing is also quite hard to teach. There are just so many things that you build up unconsciously day after day of doing something.
- When it comes to mentoring and management being more important. Mentoring: yes, management: maybe not! Truly great employees need a lot less managing by definition, and so we may just shift to prioritising autonomy and need for less management in our hiring. However coming back to your comment about professionals from outside the movement needing to rapidly get context, I do think the mentoring aspect does become very important.
- I think your point about "Engagement without employment may become core infrastructure" Might be the most important point. I don't think it's inherently bad that we shift towards this. However we must set expectations of the people doing these engagements. For example, hackathons in Tech are a great way of doing this: everyone gets together for a day, a weekend, form teams and attempt to solve a problem. After the hackathon there is no expectation that they will do a part II, or get a job working with any of the organisations they help. It's just seen as a one-off project. On the other hand, I could easily imagine that many people start fellowships and cohorts with the main priority of using it as a stepping stone to get a job in the movement. We should be very clear if our engagements are not going to help people get jobs in the movement, otherwise we might end up with lots of people becoming disillusioned.
- On your tradeoff: "Explicitly weighing internal training against external recruitment": I strongly external recruitment in all but one case. Internal training costs money we don't have, so we should be very careful about doing it. I think the essential thing to bring it back to is the idea that the tacit knowledge built up over years is the valuable thing we want. External hiring is in effect paying money for the crystallisation of someone's life experience. The only time when I think internal training makes sense is when you have activists with very deep context and experience and you train to upskill them in AI tooling. I don't think the movement should be devoting time or resources to training people who don't have useful experience and context either inside the movement or outside.
I think that your points 2 (Separating learning roles from output roles.) and 3 (Treating coordination and mentorship as core infrastructure) are important, and we've kind of beeing doing this in EA for a while as there are many orgs dedicated to fellowships and cohorts. However coming back to my point about setting expectations: many of these organisations do emphasise the job prospects of doing their cohorts. These organisations should probably be thinking about whether continuing to communicate this is a good idea or not.
Your online shopping delivery is batched with other houses' deliveries which means that the petrol required to get groceries from the supermarket to your house is lower. Also online shopping encourages shopping in bulk which means fewer trips to the supermarket.
Genuinely curious: who here would class themselves as an EA animal advocate would agree that "the political will for a phase-out will never materialise, and that we’ll be forever optimising suffering within a system we’ll never dismantle." ?
Like, it might be my experience, but I cannot ever hearing an EA animal advocate express this idea. But I could be wrong?
I agree that we know very little about nutrition, but I don't see why that should bias us against veganism.
I can't really think of a single thing that we eat in our modern diet that remotely resembles what our ancestors ate. We didn't eat chickens. We ate fruit, but none of the fruits resembled what we have now. Literally everything in our food system has been radically transformed in the last few hundred years.
So sure, we don't know for sure that veganism is a good diet for us, but doesn't your argument lead to the conclusions that we don't know if any particular food is good for us? If so, this argument shouldn't bias us towards or against any particular diet.
"I'm comfortable with pursuing protests through PauseAI US because they were a missing mood in the AI Safety discussion."
This raises and important point: if the success and impact of protests dependent on awareness of an issue / the existence of prior protests ?
We might imagine that the first protests on an issue are more impactful, and if protests reach a point where there's dozens going on regularly, that might also be uniquely effective.
If one publishes a research report, to what degree do we think that its a good idea to create a custom landing page for it, in addition to posting it on an org's site ? Would that help it show up in AI chats?
I propose this because with tools like Replit, Bolt, Lovable and the like, creating attractive landing pages for just about anything is now trivial.
"If I donate $10,000 to save two lives, I’m morally justified in taking one life because it’s convenient or enjoyable."
I see your point here, but it would more accurate to say (in the vegan offsetting case)
"If I donate $20 to save hundreds of lives, I’m morally justified in taking a dozen lives because it’s convenient or enjoyable."
Whilst the logic does change, the magnitude does, and I feel like that's important here.
"Seeing someone willingly make personal sacrifices for the sake of morality encourages others to act ethically, too."
This makes sense in theory, I am not sure it's correct in the case of veganism. Do many vegans here actually experience this? I definitely don't. I'm willing to bet most vegans experience orders of magnitude more stigma, defensiveness and abuse than moral admiration.
Looks interesting, but I am actually not fully sure what this is? A hackathon? A conference with presentations of the projects?
Cross posting my comments on your blog:
Thanks so much Sofia for building on my post! So good to see a movement leader weighing in. Great observations. To build on yours in turn:
- Love the idea of "context-absorption infrastructure". Interestingly this is something I recommend that organisations do for the opposite reason as well. I think that organisations should be trying to document as much of their context as possible because giving that context to AI allows AI to do better work. I like that this could also have the side benefit of creating a very good set of learning materials to allow new hires to rapidly come up to speed.
- I also think it's an excellent point that the movement might collapse to a smaller number of very experienced people. Something that AI can't really replicate is the tacit experience that comes from having done something every day. That kind of thing is also quite hard to teach. There are just so many things that you build up unconsciously day after day of doing something.
- When it comes to mentoring and management being more important. Mentoring: yes, management: maybe not! Truly great employees need a lot less managing by definition, and so we may just shift to prioritising autonomy and need for less management in our hiring. However coming back to your comment about professionals from outside the movement needing to rapidly get context, I do think the mentoring aspect does become very important.
- I think your point about "Engagement without employment may become core infrastructure" Might be the most important point. I don't think it's inherently bad that we shift towards this. However we must set expectations of the people doing these engagements. For example, hackathons in Tech are a great way of doing this: everyone gets together for a day, a weekend, form teams and attempt to solve a problem. After the hackathon there is no expectation that they will do a part II, or get a job working with any of the organisations they help. It's just seen as a one-off project. On the other hand, I could easily imagine that many people start fellowships and cohorts with the main priority of using it as a stepping stone to get a job in the movement. We should be very clear if our engagements are not going to help people get jobs in the movement, otherwise we might end up with lots of people becoming disillusioned.
- On your tradeoff: "Explicitly weighing internal training against external recruitment": I strongly external recruitment in all but one case. Internal training costs money we don't have, so we should be very careful about doing it. I think the essential thing to bring it back to is the idea that the tacit knowledge built up over years is the valuable thing we want. External hiring is in effect paying money for the crystallisation of someone's life experience. The only time when I think internal training makes sense is when you have activists with very deep context and experience and you train to upskill them in AI tooling. I don't think the movement should be devoting time or resources to training people who don't have useful experience and context either inside the movement or outside.
I think that your points 2 (Separating learning roles from output roles.) and 3 (Treating coordination and mentorship as core infrastructure) are important, and we've kind of beeing doing this in EA for a while as there are many orgs dedicated to fellowships and cohorts. However coming back to my point about setting expectations: many of these organisations do emphasise the job prospects of doing their cohorts. These organisations should probably be thinking about whether continuing to communicate this is a good idea or not.
Interesting point!
Your online shopping delivery is batched with other houses' deliveries which means that the petrol required to get groceries from the supermarket to your house is lower. Also online shopping encourages shopping in bulk which means fewer trips to the supermarket.
Ah cool, thanks for that context!
That is....a giant picture of my head haha But thanks for featuring me!
Genuinely curious: who here would class themselves as an EA animal advocate would agree that "the political will for a phase-out will never materialise, and that we’ll be forever optimising suffering within a system we’ll never dismantle." ?
Like, it might be my experience, but I cannot ever hearing an EA animal advocate express this idea. But I could be wrong?
I agree that we know very little about nutrition, but I don't see why that should bias us against veganism.
I can't really think of a single thing that we eat in our modern diet that remotely resembles what our ancestors ate. We didn't eat chickens. We ate fruit, but none of the fruits resembled what we have now. Literally everything in our food system has been radically transformed in the last few hundred years.
So sure, we don't know for sure that veganism is a good diet for us, but doesn't your argument lead to the conclusions that we don't know if any particular food is good for us? If so, this argument shouldn't bias us towards or against any particular diet.
"I'm comfortable with pursuing protests through PauseAI US because they were a missing mood in the AI Safety discussion."
This raises and important point: if the success and impact of protests dependent on awareness of an issue / the existence of prior protests ?
We might imagine that the first protests on an issue are more impactful, and if protests reach a point where there's dozens going on regularly, that might also be uniquely effective.
Love this post, super useful!
If one publishes a research report, to what degree do we think that its a good idea to create a custom landing page for it, in addition to posting it on an org's site ? Would that help it show up in AI chats?
I propose this because with tools like Replit, Bolt, Lovable and the like, creating attractive landing pages for just about anything is now trivial.
What would you put on such a landing page?
"If I donate $10,000 to save two lives, I’m morally justified in taking one life because it’s convenient or enjoyable."
I see your point here, but it would more accurate to say (in the vegan offsetting case)
"If I donate $20 to save hundreds of lives, I’m morally justified in taking a dozen lives because it’s convenient or enjoyable."
Whilst the logic does change, the magnitude does, and I feel like that's important here.
"Seeing someone willingly make personal sacrifices for the sake of morality encourages others to act ethically, too."
This makes sense in theory, I am not sure it's correct in the case of veganism. Do many vegans here actually experience this? I definitely don't. I'm willing to bet most vegans experience orders of magnitude more stigma, defensiveness and abuse than moral admiration.