This may be an info hazard, so consider only reading it when others have had a chance to reply and correct my mistakes.
But I don’t think it is an info hazard. My intuition is rather that it’s a very common consideration that already leads many people not to apply to EA jobs in the current environment. Also note that it only applies to EA jobs. I suppose that this article is not only about jobs at EAish organizations.
Tl;dr: A slightly adapted version of the replaceability argument that has gone out of fashion ~ 5 years ago. The quick version of my algorithm goes: If you’re unusually likely to survive, stay motivated, do something impactful without the job, don’t apply; otherwise apply.
I’ve been doing charity stuff for about 12 years now and don’t feel a decrease in motivation even though there’ve been stretches when I was almost alone in my mission. (Though I have become more chill, but I don’t know if that’s a good or a bad thing.) I don’t plan to have kids; I’m sort of financially secure; I have health insurance; I don’t have any debilitating psychological problems; etc. Most of the psychological and environmental factors that lead or force people to turn away from EA are well-aligned in my case.
That seems rare to me. Financial problems are a big reason among my friends to turn away from EA or at least to take a few years break to earn money. (Health and psychological problems are also very common among my friends, but those don’t influence this argument.)
At the same time I have a really high opinion of all these friends. Maybe half of them are probably smarter than me or otherwise more qualified or better suited for various jobs.
So if I applied for EA jobs, I guess the probability that I’d get rejected is high, but that’s really the only upside of applying. The worst thing that could happen is that I get an offer, because if I take it, one other person (or a bit less than one person in expectation) is not getting the job. And that person is more likely than me to depend on the structure and the social environment of the workplace for their psychological health and motivation, and more likely than me to depend on the salary to be able to maintain their altruistic lifestyle.
So the person I would replace at an EA job is more likely than me to either give up on EA or be forced back into the industry for money reasons. Worse: If I apply for jobs in the AI safety space, I’m likely to replace people who have a high risk of going to AI capabilities labs to have a vastly negative impact.
Then there’s also the problem that interview processes are said to be fairly unreliable. So even if the person I replace goes on to find another high-impact job where they don’t replace anyone in a bad way, maybe the interview process was wrong and I’m actually worse at the job than they would’ve been. So I’d be spending a lot of my lifetime on this job while having a negative impact. Even if the interview process was correct and I was the best applicant, maybe I was the best applicant by a margin of 5%. So even with all these rosy assumptions, I’m still just generating a minimal positive impact. (But really I can’t imagine that I’d be in the 95th percentile or so among the sorts of awesome people who apply for EA jobs. I would assume such an extreme evaluation to be a fluke just on the basis of its extremeness.)
Conversely, I can look at all the awesome people who are doing AI safety research. I can wonder who of them would be honing some potentially dangerous AI if I had applied to enough jobs to have replaced them so that they reached the end of their runway and had to fall back on their career plan Z to pay the bills.
But there are also jobs that create capacity. I imagine ops people, including CEOs and other managers, create more capacity than they absorb. So that’s probably more robustly good. Hence why I’m excited about charity entrepreneurship.
A simpler version of this replaceability argument used to be very common. But then 80k put out some articles that argued against it in the context of harmful jobs, replacing people who go on to do even greater things, etc., and it was forgotten about. But I never considered working for a tobacco company anyway, and the people I see being replaced don’t go on to do great things elsewhere but rather run out of money and end up in dead-end industry jobs. So in my mind, the old argument still applies as it always did.
What am I not getting here?
Personal note: I have a ton of anxiety around rejection. The job rejections I’ve gotten 5+ years ago still fill me with shame and I wish I could hide them from the world, make everyone forget about them, or erase them from the timelines. (Though I don’t endorse that impulse.) So it’s easily imaginable I’m biased about this. Then again I did manage to power through at the time and applied for jobs because they were ETG jobs where I don’t run into the above risks. If I had the conviction that it’s the right thing to do, I think I could’ve pushed through that barrier again. So I think the key question really is, What am I not getting here?
I think that this isn’t a useful way of looking at the situation and doesn’t match the reality well. I don’t have time to fully elaborate on why I think that, but here are brief points:
The difference between the first and second choice applicant in terms of their fit for the role can often be quite large (in expectation)
The person who would’ve been picked if the top ranked person didn’t apply or turned the job down can still probably go do something else.
This in fact very much happens; quite often the people who nearly get an offer also get another cool offer at a similar time, and that other offer is probably a better fit for them.
Also sometimes an org will help “near-miss candidates” find other opportunities (e.g., recommending them for things)
People should probably be applying to many things, which helps with this
A decent fraction of the time, an org just won’t hire someone at all if no applicant meets some high bar, or will hire an extra person if an extra person meets the high bar
I’ve observed both
My understanding is that unstructured interviews don’t have much validity or reliability, but structured interviews are better, and that will almost always be only one step of a multi-step process anyway (maybe alongside 2-3 work tests, a bit of weight on a CV, and reference checks). The process as a whole still has noise and probably bias (e.g., maybe screening harder than is ideal for people with prior EA or cause-area knowledge), but it’s probably far more reliable and valid than guesswork, and reliable and valid enough to still create substantial expected differences between the impact of the top ranked people taking the role and the impact of the next ranked people taking the role.
Even “individual contributor” roles like researcher, as opposed to ops or manager type roles, can still help scale up an org or field and help new people get in. E.g., researchers could do distillation or disentanglement work that makes it easier for other people to get up to speed or find out where to contribute.
If you’re otherwise unusually likely to turn away from EAish stuff – i.e. reach the end of your runway or burn out – just apply. Probably even if you’re just at an average level of risk.
If you can see yourself turning down an awesome offer because you disagree with the result of the interview process, apply a bit more liberally than otherwise.
When prioritizing between positions, assign:
1000x weight to completely idiosyncratic high-impact projects (regardless of whether they’re your ideas or someone else’s) that no one else would otherwise pursue for a long time,
100x weight to relatively neglected roles in the community (say, because they require a rare combination of skills or because the org is new and fairly unknown),
10x to any capacity-creating kind of role to reduce the risk that they may not find anyone, and
Thank you for taking the time to write up the summary!
Possibly. I’ve only hired for two roles so far (using a structured process). In one case there were clear candidates 1, 2, and 3+, and while 2 might’ve just had a bad day, we made offers to 1 and 2 anyway. In another case, though, we had two, or possibly three, candidates tied for the top spot. Two, we thought, would be more pleasant to work with while the third one seemed to have the stronger technical skill. We didn’t know how to trade that off and ended up making the offer to the one with the stronger technical skill. I have no idea whether that was the right call.
Yes, that’s helpful for mitigating the worst-case risks. We also did that in the second case. It still seems weak though. I imagine that in most cases they’re not able to help the other candidates very much. We weren’t either afaik.
Yes, that’s also a system I’ve encountered, and I love it! That’s a strong reason in my mind to apply somewhere after all. But I don’t fully trust it.
Even if an organization has enough funding for this system, they may not have enough management capacity.
They may still have a hiring goal, and upon reaching it will wind down the effort they put into hiring. That frees up resources at the org at the expense of missing out on an even better candidate. The hiring process is hopefully short in comparison to the time that the person will stay at the org, so the second probably has more leverage.
I’d be replacing .5 or .2 people, which is much better, but no where near an ops job that creates capacity.
Okay, that’s reassuring, but see my point 1. Then again most EA interview processes (e.g., the CLR one that Stefan described in detail a few years back) are more sophisticated than ours was. A good interview process is another minor but valuable mitigation in my mind.
Good point, but these jobs are probably in higher demand than ops jobs in EA, so the counterfactual effect is milder.
I think one crux might be that if I want to dedicate the next 5–10 years of my life to something, I have a higher bar than just “We’ve taken several precautions to make it less likely that you’ll have a vastly negative effect with your work.” Those precautions are invaluable of course, but there are better alternatives for applicants.
It also take a very particular kind of mental fortitude to apply for 20+ roles, and when eventually you do get an offer, to turn it down because you think you’re likely not the best candidate. That seems like such a hard decision to make, especially if the job is really awesome, fun, and high status.
The last 18 months I have applied to 9+- roles at a company I am very interested in and received rejections for 6 roles. 4 roles are under review. I reached out to a recruiter and he said the hiring managers expressed no interest to interview me. Should I stop applying at that company? I am afraid I may be annoying the recruiters if continue to apply to their jobs after so many rejections. What do you think?
Hi Richard, quick reactions without having much context:
If you mean this is all one company, this sounds like putting too many eggs in one basket, and insufficiently exploring.
I think it’s generally good to apply to many different types of roles and organizations.
Sometimes it makes sense to focus in mostly on one role type or one org. But probably not entirely. And not once one has already gotten some evidence that that’s not the right fit. (Receiving a few rejections isn’t much negative info, but if it’s >5 for one particular org or type of thing then that’s probably at least enough evidence that one should also apply to lots of other things and not spend lots of further time on this one thing.)
I’d be much less focused on “am I annoying them?” than “Am I spending too much of my valuable time on this one type of thing, and also potentially missing lots of other better-fitting things elsewhere?”
This may be an info hazard, so consider only reading it when others have had a chance to reply and correct my mistakes.
But I don’t think it is an info hazard. My intuition is rather that it’s a very common consideration that already leads many people not to apply to EA jobs in the current environment. Also note that it only applies to EA jobs. I suppose that this article is not only about jobs at EAish organizations.
Tl;dr: A slightly adapted version of the replaceability argument that has gone out of fashion ~ 5 years ago. The quick version of my algorithm goes: If you’re unusually likely to survive, stay motivated, do something impactful without the job, don’t apply; otherwise apply.
I’ve been doing charity stuff for about 12 years now and don’t feel a decrease in motivation even though there’ve been stretches when I was almost alone in my mission. (Though I have become more chill, but I don’t know if that’s a good or a bad thing.) I don’t plan to have kids; I’m sort of financially secure; I have health insurance; I don’t have any debilitating psychological problems; etc. Most of the psychological and environmental factors that lead or force people to turn away from EA are well-aligned in my case.
That seems rare to me. Financial problems are a big reason among my friends to turn away from EA or at least to take a few years break to earn money. (Health and psychological problems are also very common among my friends, but those don’t influence this argument.)
At the same time I have a really high opinion of all these friends. Maybe half of them are probably smarter than me or otherwise more qualified or better suited for various jobs.
So if I applied for EA jobs, I guess the probability that I’d get rejected is high, but that’s really the only upside of applying. The worst thing that could happen is that I get an offer, because if I take it, one other person (or a bit less than one person in expectation) is not getting the job. And that person is more likely than me to depend on the structure and the social environment of the workplace for their psychological health and motivation, and more likely than me to depend on the salary to be able to maintain their altruistic lifestyle.
So the person I would replace at an EA job is more likely than me to either give up on EA or be forced back into the industry for money reasons. Worse: If I apply for jobs in the AI safety space, I’m likely to replace people who have a high risk of going to AI capabilities labs to have a vastly negative impact.
Then there’s also the problem that interview processes are said to be fairly unreliable. So even if the person I replace goes on to find another high-impact job where they don’t replace anyone in a bad way, maybe the interview process was wrong and I’m actually worse at the job than they would’ve been. So I’d be spending a lot of my lifetime on this job while having a negative impact. Even if the interview process was correct and I was the best applicant, maybe I was the best applicant by a margin of 5%. So even with all these rosy assumptions, I’m still just generating a minimal positive impact. (But really I can’t imagine that I’d be in the 95th percentile or so among the sorts of awesome people who apply for EA jobs. I would assume such an extreme evaluation to be a fluke just on the basis of its extremeness.)
Conversely, I can look at all the awesome people who are doing AI safety research. I can wonder who of them would be honing some potentially dangerous AI if I had applied to enough jobs to have replaced them so that they reached the end of their runway and had to fall back on their career plan Z to pay the bills.
But there are also jobs that create capacity. I imagine ops people, including CEOs and other managers, create more capacity than they absorb. So that’s probably more robustly good. Hence why I’m excited about charity entrepreneurship.
A simpler version of this replaceability argument used to be very common. But then 80k put out some articles that argued against it in the context of harmful jobs, replacing people who go on to do even greater things, etc., and it was forgotten about. But I never considered working for a tobacco company anyway, and the people I see being replaced don’t go on to do great things elsewhere but rather run out of money and end up in dead-end industry jobs. So in my mind, the old argument still applies as it always did.
What am I not getting here?
Personal note: I have a ton of anxiety around rejection. The job rejections I’ve gotten 5+ years ago still fill me with shame and I wish I could hide them from the world, make everyone forget about them, or erase them from the timelines. (Though I don’t endorse that impulse.) So it’s easily imaginable I’m biased about this. Then again I did manage to power through at the time and applied for jobs because they were ETG jobs where I don’t run into the above risks. If I had the conviction that it’s the right thing to do, I think I could’ve pushed through that barrier again. So I think the key question really is, What am I not getting here?
I think that this isn’t a useful way of looking at the situation and doesn’t match the reality well. I don’t have time to fully elaborate on why I think that, but here are brief points:
The difference between the first and second choice applicant in terms of their fit for the role can often be quite large (in expectation)
The person who would’ve been picked if the top ranked person didn’t apply or turned the job down can still probably go do something else.
This in fact very much happens; quite often the people who nearly get an offer also get another cool offer at a similar time, and that other offer is probably a better fit for them.
Also sometimes an org will help “near-miss candidates” find other opportunities (e.g., recommending them for things)
People should probably be applying to many things, which helps with this
A decent fraction of the time, an org just won’t hire someone at all if no applicant meets some high bar, or will hire an extra person if an extra person meets the high bar
I’ve observed both
My understanding is that unstructured interviews don’t have much validity or reliability, but structured interviews are better, and that will almost always be only one step of a multi-step process anyway (maybe alongside 2-3 work tests, a bit of weight on a CV, and reference checks). The process as a whole still has noise and probably bias (e.g., maybe screening harder than is ideal for people with prior EA or cause-area knowledge), but it’s probably far more reliable and valid than guesswork, and reliable and valid enough to still create substantial expected differences between the impact of the top ranked people taking the role and the impact of the next ranked people taking the role.
Even “individual contributor” roles like researcher, as opposed to ops or manager type roles, can still help scale up an org or field and help new people get in. E.g., researchers could do distillation or disentanglement work that makes it easier for other people to get up to speed or find out where to contribute.
An alternative decision algorithm:
If you’re otherwise unusually likely to turn away from EAish stuff – i.e. reach the end of your runway or burn out – just apply. Probably even if you’re just at an average level of risk.
If you can see yourself turning down an awesome offer because you disagree with the result of the interview process, apply a bit more liberally than otherwise.
When prioritizing between positions, assign:
1000x weight to completely idiosyncratic high-impact projects (regardless of whether they’re your ideas or someone else’s) that no one else would otherwise pursue for a long time,
100x weight to relatively neglected roles in the community (say, because they require a rare combination of skills or because the org is new and fairly unknown),
10x to any capacity-creating kind of role to reduce the risk that they may not find anyone, and
1x to any other role.
Thank you for taking the time to write up the summary!
Possibly. I’ve only hired for two roles so far (using a structured process). In one case there were clear candidates 1, 2, and 3+, and while 2 might’ve just had a bad day, we made offers to 1 and 2 anyway. In another case, though, we had two, or possibly three, candidates tied for the top spot. Two, we thought, would be more pleasant to work with while the third one seemed to have the stronger technical skill. We didn’t know how to trade that off and ended up making the offer to the one with the stronger technical skill. I have no idea whether that was the right call.
Yes, that’s helpful for mitigating the worst-case risks. We also did that in the second case. It still seems weak though. I imagine that in most cases they’re not able to help the other candidates very much. We weren’t either afaik.
Yes, that’s also a system I’ve encountered, and I love it! That’s a strong reason in my mind to apply somewhere after all. But I don’t fully trust it.
Even if an organization has enough funding for this system, they may not have enough management capacity.
They may still have a hiring goal, and upon reaching it will wind down the effort they put into hiring. That frees up resources at the org at the expense of missing out on an even better candidate. The hiring process is hopefully short in comparison to the time that the person will stay at the org, so the second probably has more leverage.
I’d be replacing .5 or .2 people, which is much better, but no where near an ops job that creates capacity.
Okay, that’s reassuring, but see my point 1. Then again most EA interview processes (e.g., the CLR one that Stefan described in detail a few years back) are more sophisticated than ours was. A good interview process is another minor but valuable mitigation in my mind.
Good point, but these jobs are probably in higher demand than ops jobs in EA, so the counterfactual effect is milder.
I think one crux might be that if I want to dedicate the next 5–10 years of my life to something, I have a higher bar than just “We’ve taken several precautions to make it less likely that you’ll have a vastly negative effect with your work.” Those precautions are invaluable of course, but there are better alternatives for applicants.
It also take a very particular kind of mental fortitude to apply for 20+ roles, and when eventually you do get an offer, to turn it down because you think you’re likely not the best candidate. That seems like such a hard decision to make, especially if the job is really awesome, fun, and high status.
The last 18 months I have applied to 9+- roles at a company I am very interested in and received rejections for 6 roles. 4 roles are under review. I reached out to a recruiter and he said the hiring managers expressed no interest to interview me. Should I stop applying at that company? I am afraid I may be annoying the recruiters if continue to apply to their jobs after so many rejections. What do you think?
Hi Richard, quick reactions without having much context:
If you mean this is all one company, this sounds like putting too many eggs in one basket, and insufficiently exploring.
I think it’s generally good to apply to many different types of roles and organizations.
Sometimes it makes sense to focus in mostly on one role type or one org. But probably not entirely. And not once one has already gotten some evidence that that’s not the right fit. (Receiving a few rejections isn’t much negative info, but if it’s >5 for one particular org or type of thing then that’s probably at least enough evidence that one should also apply to lots of other things and not spend lots of further time on this one thing.)
I’d be much less focused on “am I annoying them?” than “Am I spending too much of my valuable time on this one type of thing, and also potentially missing lots of other better-fitting things elsewhere?”
One company. You are right about too many eggs in one basket. I’m expanding my search to more companies and focusing on operations roles.
I learned recently my resume is too generic, not targeted enough to the roles and needs quantifiable accomplishments.
I’m updating...Thank you.