In most cases, especially high-stakes cases, cost of living is small relative to prioritization/productivity/information benefits. (View not justified here.)
It seems pretty likely to me that you instead want AIs to be risk seeking for reasons discussed here. Takeover attempts that are very unlikely to succeed might speculatively actually be a great trade from the perspective of humanity and risk seeking/risk neutral AI in that they reduce overall takeover risk while being good for this AI (while deals are way less useful to humanity due to being less clear evidence). Risk avoidant AIs might do nothing or just take deals until AIs take over later (and accepting deals might not be a good strategy for them depending on their views about the chance of AI takeover and some other factors).
I also think the implicit story about how we steer these traits doesn't really hold together and assumes a type of generalization I find somewhat implausible if we condition on AIs being egregiously misaligned.
If AGI goes well for humans, it’ll probably [i.e. ≥70%] go well for animals
I think it doesn't really make sense to do a sliding-scale vote on ≥70%. If my credence on (AGI goes well for animals | AGI goes well for humans) is 60%, then I'm just a no; if my credence is 80%, then I'm just a yes.
One way people could interpret the sliding-scale is expressing their confidence/stability in their judgment about whether the probability is greater or less than 70%. But that's somewhat deranged and it's not clear how to make it precise and I think everyone will just be confused.
It would be totally reasonable to vote on a scale from 0% to 100% for P(AGI goes well for animals | AGI goes well for humans), rather than voting from fully-disagree to fully-agree for ≥70%. Obviously that requires a little recoding. But making the voting scale more flexible, rather than just from fully-disagree to fully-agree, will benefit other debates too.
My claim is that if you're worried, the correct response is to actually try to make the astronomical problem/cause go better, not to give up on it. I think if you're savvy you will probably find a way to make the astronomical thing go better—such as doing strategy/prioritization/deconfusion work, or working on robustly good intermediate desiderata, or building skills/money in case there's more clarity in the future—rather than ultimately thinking there's nothing I can do to make the thing go better.
Once upon a time, there was an EA named Alice. EA made a lot of sense to Alice, and she believed that some niche problems/causes were astronomically bigger than others. But she eventually decided that (1) the theories of change were confusing/suspicious and (2) there's substantial evidence that a bunch of EA work is net-negative. So she decided to become a teacher or doctor or something.
II.
Alice made a mistake! If she thinks that some problems/causes are astronomically bigger than others, and she's skeptical of certain approaches, she should look for better approaches, not give up on those problems/causes! For example, she could:
Find an intervention (in the great problems/causes) that she believes in, and do that
Defer to people who she really respects on the topic
Try to understand the problem and possible interventions; do strategy/prioritization/deconfusion work (for herself or maybe benefitting the whole community)
Develop relevant skills and/or save up money, and set herself up to notice if there's more clarity or great opportunities in the future
Accept sign-uncertainty and do positive-EV stuff
III.
This is actually about my friend Bob who's sometimes like I work on AI safety but I feel clueless about whether we're actually helping, and I see that farmed animal suffering is a huge problem, and I want to go work on farmed animal welfare. If Bob still believes that the AI stuff is astronomically more important than the animal stuff, Bob is making the same mistake as Alice!
No, funders can get even more money for effective philanthropy by investing in AI.
In most cases, especially high-stakes cases, cost of living is small relative to prioritization/productivity/information benefits. (View not justified here.)
Crossposting Ryan's comment on LW:
Prior art: this, this, and some of this.
Over on LessWrong I wrote some posts about prioritization research and donations.
Yes. Most people will directionally-agree (which is maybe a problem you were trying to solve by adding "probably") but maybe that's OK.
I think it doesn't really make sense to do a sliding-scale vote on ≥70%. If my credence on (AGI goes well for animals | AGI goes well for humans) is 60%, then I'm just a no; if my credence is 80%, then I'm just a yes.
One way people could interpret the sliding-scale is expressing their confidence/stability in their judgment about whether the probability is greater or less than 70%. But that's somewhat deranged and it's not clear how to make it precise and I think everyone will just be confused.
It would be totally reasonable to vote on a scale from 0% to 100% for P(AGI goes well for animals | AGI goes well for humans), rather than voting from fully-disagree to fully-agree for ≥70%. Obviously that requires a little recoding. But making the voting scale more flexible, rather than just from fully-disagree to fully-agree, will benefit other debates too.
I haven't engaged with your posts and so don't know the arguments.
I respect that you and a few others legitimately feel deeply clueless. Alice and Bob are just whining about how not everything is clear-cut.
My claim is that if you're worried, the correct response is to actually try to make the astronomical problem/cause go better, not to give up on it. I think if you're savvy you will probably find a way to make the astronomical thing go better—such as doing strategy/prioritization/deconfusion work, or working on robustly good intermediate desiderata, or building skills/money in case there's more clarity in the future—rather than ultimately thinking there's nothing I can do to make the thing go better.
I.
Once upon a time, there was an EA named Alice. EA made a lot of sense to Alice, and she believed that some niche problems/causes were astronomically bigger than others. But she eventually decided that (1) the theories of change were confusing/suspicious and (2) there's substantial evidence that a bunch of EA work is net-negative. So she decided to become a teacher or doctor or something.
II.
Alice made a mistake! If she thinks that some problems/causes are astronomically bigger than others, and she's skeptical of certain approaches, she should look for better approaches, not give up on those problems/causes! For example, she could:
III.
This is actually about my friend Bob who's sometimes like I work on AI safety but I feel clueless about whether we're actually helping, and I see that farmed animal suffering is a huge problem, and I want to go work on farmed animal welfare. If Bob still believes that the AI stuff is astronomically more important than the animal stuff, Bob is making the same mistake as Alice!