Alex Lawsen on avoiding 10 mistakes people make when pursuing a high-impact career

By , , and · Published September 2023

See all episodes

Enjoyed the episode? Want to listen later? Subscribe here, or anywhere you get podcasts:

I think the key consideration that I end up highlighting to people who I think are trying to do the best thing right now is something like: It might be that setting yourself up well to do more good later looks like not directly having as much of an impact right now.

Because probably learning is pretty good if you want to have an impact later. Probably getting some signalling experience for a lot of careers, maybe doing a prestigious internship, maybe getting paid a lot of money: these things just often pay off later, and often trade off against doing the very most good with the summer internship you’re doing this year, or in the first two years of your job just after you’ve graduated.

Alex Lawsen

In this episode of 80k After Hours, Luisa Rodriguez and Alex Lawsen discuss common mistakes people make when trying to do good with their careers, and advice on how to avoid them.

They cover:

  • Taking 80,000 Hours’ rankings too seriously
  • Not trying hard enough to fail
  • Feeling like you need to optimise for having the most impact now
  • Feeling like you need to work directly on AI immediately
  • Not taking a role because you think you’ll be replaceable
  • Constantly considering other career options
  • Overthinking or over-optimising career choices
  • Being unwilling to think things through for yourself
  • Ignoring conventional career wisdom
  • Doing community work even if you’re not suited to it

Who this episode is for:

  • People who want to pursue a high-impact career
  • People wondering how much AI progress should change their plans
  • People who take 80,000 Hours’ career advice seriously

Who this episode isn’t for:

  • People not taking 80k’s career advice seriously enough
  • People who’ve never made any career mistakes
  • People who don’t want to hear Alex say “I said a bunch of stuff, maybe some of it’s true” every time he’s on the podcast

Get this episode by subscribing to our more experimental podcast on the world’s most pressing problems and how to solve them: type ’80k After Hours’ into your podcasting app. Or read the transcript below.

Producer and editor: Keiran Harris
Audio Engineering Lead: Ben Cordell
Technical editing: Milo McGuire and Dominic Armstrong
Additional content editing: Luisa Rodriguez and Katy Moore
Transcriptions: Katy Moore

“Gershwin – Rhapsody in Blue, original 1924 version” by Jason Weinberger is licensed under creative commons

Highlights

Why you shouldn't "just do whatever 80k says is best"

Luisa Rodriguez: One thing you hear a lot is something like, “I should just do the thing that 80k says is best.” Can you say more about what that looks like?

Alex Lawsen: Yeah, I think there are a bunch of different ways this can come up. Maybe the most obvious one is just that, at any particular time, there’s going to be something really salient that it seems like lots of people should do. And maybe it’s because it’s at the top of the ranked list we have. That is one disadvantage of having ranked lists. So a very clean version of this mistake could just be people going, “The thing that’s at number one of the list of career profiles on the 80k website is technical AI safety research, so I guess I have to do technical AI safety research.”

So why might this be a mistake? The obvious thing is just that personal fit really matters. It matters in a bunch of ways. My guess is, during a lot of this conversation, I’m going to end up saying, “Did you know that personal fit is a big deal?” But specifically in the case of someone correctly realising that a thing is really important for more people to do, and then still making a mistake by thinking that they have to do it, what’s going on is something like it’s just a really bad fit for them. And I think there’s one way which is easy for people to internalise this, which is like, “I wouldn’t be very good at this, because I’m dumb or stupid” or something.

And I always say that people are really different. People are good at different things. And not being able to do the thing that’s at the number one place, or the thing that your friend who’s read more of the website than you says is the best thing… Maybe there’s some sense where if you realise that the thing that’s going on there is you aren’t best placed to do that thing — maybe because you would hate it; maybe you’d be good at it, but you would just hate it and you would not manage to force yourself to do it for more than a couple of years — just realising, “That’s not the best thing for me to do. I should do something else.”

I have some guess that it’s easier to realise that maybe you should do something else if you realise that one of the things that you should be tracking is who is best placed to work on what. I can imagine some weird world where you’ve got this ranking of the best things for people to do, and then you just rank people by ability — but the distribution of ability is exactly the same across all the people: you just get like the best people and you put them in the best thing. Once that thing’s full up, you get the next best people, put them the next best thing. The world just doesn’t work like that. That’s maybe the thing that seems top of mind, is this: “I’ve heard one job is really important and I have to do that.”

Trying harder to fail

Alex Lawsen: I’m pretty sure I mentioned in the only other interview I’ve done on the 80k podcast was about how I approached exams when I was a teenager, and actually in university as well, which was really not preparing as much as my classmates would. And there were some other things going on that affected this, but certainly my internal narrative was like, “If I do well, having not tried that hard, then there’s nothing ruling out the possibility that I could have done astonishingly well if I had tried really hard.”

And that internal narrative is exactly the problem. The thing you want to do is rule out the possibility that this is the perfect match for you, so that you can consider whether something else is. And if you keep yourself in this limbo of, like, “Maybe this would be amazing if I tried harder at it…” Yeah. There’s something about, like, failing is really scary. And then it can also just free you up to try stuff that’s just much better for you and for everyone else.

Luisa Rodriguez: Right, yeah. Like, “Great, I have one fewer option for things. There are like a billion things I could do with my career. I don’t have to worry about that one.” Which is probably just good if you’re not feeling like you’re crushing it and enjoying it. Are there other things that you’ve seen as an advisor, or other ways this can look?

Alex Lawsen: Yeah, I think there’s a few. Often these will come up in early-career people who are trying to test out a bunch of different things. Let’s say you think you might want to be able to do some kind of writing-based role. Maybe you want to be a journalist or a public intellectual, or even just like a successful blogger. And so you write a bunch of drafts and you never really share them, or you post like one every six months. It’s on the EA Forum, but not under your own name.

And I think what’s going wrong here is just probably you’re not going to get a tonne of feedback either way, and you’re probably not going to get that much better at writing. Writing careers are hard. Forget impact: just from a personal perspective, you’re really going to have to have a bunch of stuff go well — not just in terms of talent for clarity or excision of ideas or new generative ideas, but even just how fast you can put stuff out. And yeah, if you’re just writing a little bit sometimes and then not putting it out there, you can do this for two years and still have no idea how good your writing is.

Luisa Rodriguez: Right. You get this ambiguous result. And then what do you do with an ambiguous result? You’ve got the nice feeling of like, maybe I could be crushing it, and I don’t have to stare in the face the fact that I’m unambiguously not crushing it. But yeah, you don’t know if that’s the thing you should be doing because it’s going fine, but not exceptionally well.

Alex Lawsen: And there’s some sense in which this is sort of safe in the near term, that you’re not going to hear, “Oh, wow. Try something else.” But actually, if you should try something else — because, just as a random hypothesis, it turns out you might be really good at podcasting — never getting that signal, never getting the nudge to try a different thing, ends up harming you and ends up being difficult.

Luisa Rodriguez: I feel like having success and failure criteria that are objective enough to measure yourself against seems really key here.

Alex Lawsen: Yeah. Here’s a recent example that could be kind of cool. There was a post on the EA Forum fairly recently about ways to test fit for biosecurity careers. I think it was aimed at fairly early-career people, and it was essentially just a kind of smallish research project: Can I produce a report? And it doesn’t have to be original research, but can I produce a summary of some part of this field in 10 hours — five hours a day across the weekend? And then I’ll send that to some people and see whether they think it’s nonsense. But even before I send it to anyone, you’ve got some signal on: How was it to do that? How would I feel if my week was doing that? And I think often this can just come quite quickly.

Why you shouldn't optimise for impact *now*

Luisa Rodriguez: Another common mistake that you hear is people feeling like they need to optimise for having the most impact they can right now. Why do you think that is a mistake?

Alex Lawsen: Yeah. Why is this a mistake? Let’s say that with some part of your time or effort or whatever, you want to do the most good. The thing that I claim most people care about, when they are trying to do this, actually is the most good across your life, or across the time that you’re working. Let’s say I’m a millionaire. I won the lottery. I won a million pounds, and I want to give that away. And I’m going to give away £1,000 a year for the next 1,000 years, because I just set myself up badly. I claim the thing I should be thinking about is how do I do the best with that million? — not how do I do the best with the first thousand — even at the cost of losing money. What this might look like in the money case is that maybe I should hire some researchers to help me work out what to do with the rest of the money.

And I think the key consideration that I end up highlighting to people who I think are trying to do the best thing right now is something like: It might be that setting yourself up well to do more good later looks like not directly having as much of an impact right now. Because probably learning is pretty good if you want to have an impact later. Probably getting some signalling experience for a lot of careers, maybe doing a prestigious internship, maybe getting paid a lot of money: these things just often pay off later, and often trade off against doing the very most good with, like, the summer internship you’re doing this year, or in the first two years of your job just after you’ve graduated.

Luisa Rodriguez: In my case, it feels like it was a combination of wanting to feel good about feeling and thinking that I am having a positive impact, hopefully, on the world; wanting to feel like I’m living by my values of doing the most important thing, and not a thing that seems pretty good. And then also just totally something about wanting my peers to perceive me as doing good and living by my values. So wanting to work somewhere that’s, like, legibly high impact, and not just somewhere that’s giving me career capital, but doesn’t seem as obviously good to them. And I think I’ve had this particularly because being part of the effective altruism community, there are lots of people who all share some overlapping views on what’s allowed, what’s high status — and I just felt super hyperaware of that, and wanted the approval, I guess, from people around me, of doing the good thing now. It’s just a lot of things pushing in that direction. How do we help people push in the other direction?

Alex Lawsen: I certainly think that part of what’s going on sometimes here is this peer approval thing. Maybe, actually, it’s worth saying something like: If you’re a member of this community, or adjacent to it, and you’re talking to lots of other people in a similar position, kind of like making it clear that you won’t judge your friends for taking things that seem pretty normal career capital building instead of the, you know, weird EA internship that seems really cool because it’s got like a four-letter acronym running it.

But yeah, I think part of it can be pressure. The main thing I do to push back on this in calls is just hammering the thing: What are you trying to achieve? You’re trying to achieve the most good — not the most good this year in exchange for less good later. I do think that part of the difficulty is that maybe people are really worried about suspiciously convenient outcomes. Where you like take the high-status job at the quant trading firm, where you’re going to get to play with fun maths puzzles every day. And earn a bunch of money, and be seen as super duper smart by your normal peers who don’t have any idea what this altruism stuff is. That’s a real draw. And I think if that’s a draw that you don’t endorse, because you really care about helping as many people as you can, then it’s easy to not realise that there are “helping as many people as you can”-style reasons for doing that thing anyway.

Luisa Rodriguez: I think it still just feels really costly to me to have to sacrifice the feeling of I’m living by my values, and I feel very proud of that, and it’s a thing I care about a lot. I mean I just wish it weren’t the case, but it is the fact that I care loads that my peers also think I’m living by my values. Is there more from just, like, a sociological perspective that people can do?

Alex Lawsen: Yeah. I think this is a really good thing to push on. Maybe the first thing that seems important is, if I was speaking to Luisa three years ago on an advising call, and she was saying stuff like, “I really want to live up to my values,” the main thing I’d push on is like, “Wait. Your values are about doing the best thing, not the best thing right now. That’s actually the thing you care about.” So I might want to push in the direction of, if I can convince you that this does actually seem best in expectation across your long-run career, then that is the thing that’s in accordance with your values.

And I don’t know whether I would make this next move with everyone, but I find it plausible I might even say something like, “You care about acceptance from people in this immediate social circle of do-gooders, not from the world at large. It’s actually pretty unclear to me that you’re taking the easy route out by doing the normie prestigious internship, rather than the weird self-sacrificing thing that pays you much less. Plausibly, among the people whose opinions you care about, this is the hard option. This is the option where you have to explain your choice, and you feel like it’s not something that has their EA-approved stamp on it.” And I’m like, “You want to live up to your values? You want to do the brave thing? That’s the brave thing.” Sure, the normal world won’t judge you for it. But maybe the normal world isn’t the world you’re paying attention to.

How to stop constantly considering other career options

Luisa Rodriguez: Another one that I’m super familiar with is basically being on a reasonably high-impact path or in a high-impact role, but still constantly considering other career options that might be even more impactful, in a way that’s either kind of distracting or leads to too much switching, and never getting great at one set of skills. I feel like for years I was doing this. I think I’d been working for six or seven years, and I’d never had a job for more than 12 months.

And that was basically, I’d be in a role that I thought was reasonably worth doing, seemed kind of valuable. And then some low self esteem-y thing creeped in, where I was like, “What I’m doing probably isn’t that impactful, because it’s me, and I surely couldn’t be doing anything that impactful.” If ever another opportunity was available and seemed like it might be high impact, and it made the difference between my role and that role seem artificially big to me. I was just like, “Ooh, a chance to actually have a big impact instead of do this bad thing I’m currently doing, which is not impactful because I’m bad.” It wasn’t quite that extreme, but I think that was a thing pushing in that direction.

I think there are probably other things going on too. There’s a thing that can happen where it feels interesting and exciting and enticing to potentially get to do an even more impactful thing. And so whenever something’s presented, you’re just like, “Ooh, new thing! Potential for a new, bigger impact.” And for years it just meant I ended up being in a role for a bit, and someone suggested I apply for another one. In some cases, I got those roles, and then I’d switch because of a bunch of these biases, and then spent very little time getting actually very good at one thing because I’ve done it for years or something.

Alex Lawsen: Yeah, I think there’s a difficult balance to strike here. It’s not the case that people should just pick one thing and then go, “Great, this is my thing.” And I think people make this mistake a lot too, especially if there’s some “profession” that you’re in. So I think being a teacher, being a lawyer, being a doctor feels like one of these things where, like, “This is the path I’m on. Now all of my options are things within this path.” So yeah, it is hard to strike a balance, because I think people realise this and then they go, “I want to do the best thing,” and then they’re always looking for something else.

Where does this go wrong? There are a couple of different ways. One way — and from the sound of it, this has happened to you before — is you keep switching and you don’t get that much practice. Skills are transferable, but not all of the skills. Or even just you don’t settle into your groove, you feel kind of like, “How much should I even be investing into this place? Probably I’m going to have to switch in a year when I find something better.” That’s one way this can go wrong, and I think emotionally this can be pretty hard.

Another thing, and I see this actually quite a lot too, is people are doing something that seems to me objectively great, and they’re just feeling terrible about it and not putting that much time or energy into it, because they’re doing a bunch of job hunting on the side. Because what if it’s not the thing that’s objectively great? What if there’s something better? And applying for jobs takes time, it takes energy, it takes emotional resilience. And feeling like you constantly have to be putting some of your attention on whether there’s something better, even if you don’t then switch to it, still results in you paying pretty significant costs.

Luisa Rodriguez: Yeah, that sounds totally right. So what is the better decision-making process, if this is the kind of trap you fall into?

Alex Lawsen: I have a thing which has worked really well for me. I think it works well for lots of people, and probably won’t work well for everyone. I think I’ll just try to describe the thing, but then I’ve got a metaphor I’ve been playing with that maybe I want to try and describe as well. So the thing is, you stick to some policy, which is like: “I’m going to look at a bunch of things, I’m going to actually seriously consider my options. And then with all of the information I have, I’m going to make a decision. And I’m going to make the decision to do the thing that seems best for some fixed period of time. At the end of that fixed period of time, then I will consider other options.”

And I think for most people, you don’t have to make this a fully binding commitment, that you just instantly say no as soon as anyone says, “Have you considered…?” I think of it as a strong default. If someone says, “Hey, have you thought about doing this other thing?” I can just say, “The next time I’m looking at options is in a couple of years. If you want to pitch me on this really hard, I’m not going to rule out ever doing anything before that time, ever, but you should know that you’ve got a lot of work to do to convince me to even really think about this. Because my plan is to stay for this long.”

Going too far in a good direction

Luisa Rodriguez: Before we wrap up, I am curious if you had any reflections having done this, or any other mistakes that come to mind that we can cover before we call it?

Alex Lawsen: Maybe one theme that seems worth pulling out is something like going too far in a good direction. I think lots of the stuff we’ve talked about has been like, “Conventional wisdom says this. You’ve noticed that conventional wisdom seems to not be tracking quite the thing you care about, or maybe even you just think conventional wisdom is wrong. And so you go, ‘I want to move conventional wisdom in this direction'” — and then you just move it way too hard in that direction. I don’t want to even make the general claim that you should never move things that far. I in fact think you should push some things pretty hard, and keep other things at a satisficing level. But I think there’s some amount of the correct update to make is not always like a really, really big one.

Luisa Rodriguez: To the other extreme. Yeah. That just makes sense, and it also makes sense that it’s hard, and so lots of people make these “mistakes.” If the optimal amount to move in a direction is a bit, but not all the way to the other extreme, finding the exact spot that’s ideal is just going to be really hard. And especially when you’ve got decisions as hard as these. We’ve already kind of qualified that people should not feel bad or guilty or ashamed if they’ve done some of these things. It’s just difficult because you’re doing this balancing act.

Alex Lawsen: I think it’s actually a dynamic of how language is structured, or how people speak to each other, that makes this hard to talk about. And I don’t mean hard emotionally; I mean we don’t quite have the right vocabulary for it. The thing that I want to point out here is we don’t have a neat single word for “I agree with you in the direction, but not as hard as you.” If people are estimating probabilities, then there’s still not one word, but there’s an idea. Let’s say you think there’s a 90% chance something’s going to happen, and I think there’s a 75% chance it’s going to happen. In this case, I can say I think it’s less likely, and it’s still kind of hard to explain why. If I explain why it’s less likely, but I don’t say any numbers, my guess is an observer that didn’t hear the start of the conversation isn’t really going to be able to tell whether I’m saying 75% or 10%.

Luisa Rodriguez: Right. It’s like, are you saying it’s less likely than I think and therefore you think it’s unlikely? Or are you saying it’s less likely than I think, it’s still probably going to happen, but you believe it less strongly than I do?

Alex Lawsen: Yeah, I think that’s exactly it. It’s just kind of clumsy and awkward to talk about situations where you and the person you’re talking to both basically think this thing on the left, rather than this thing on the right that everyone else thinks, but you’re a bit closer to the right than they are. There’s just kind of a difficult dynamic. And so I think part of what’s going on here is if lots of your friends and people you talk to about your career decisions are in a kind of similar place to you compared to conventional wisdom, it’s going to be hard for them to error correct. They don’t have quite the right vocabulary to error correct without just saying conventional wisdom at you.

Articles, books, and other media discussed in the show

Alex’s work:

Career resources from 80,000 Hours:

Other 80,000 Hours podcast episodes:

Everything else:

Transcript

Cold open [00:00:00]

Alex Lawsen: If you’re just writing a little bit sometimes and then not putting it out there, you can do this for two years and still have no idea how good your writing is.

Luisa Rodriguez: Right. You get this ambiguous result. And then what do you do with an ambiguous result? You’ve got the nice feeling of like, maybe I could be crushing it, and I don’t have to stare in the face the fact that I’m unambiguously not crushing it. But yeah, you don’t know if that’s the thing you should be doing because it’s going fine, but not exceptionally well.

Alex Lawsen: And there’s some sense in which this is sort of safe in the near term, that you’re not going to hear, “Oh, wow. Try something else.” But actually, if you should try something else — because, just as a random hypothesis, it turns out you might be really good at podcasting — never getting that signal, never getting the nudge to try a different thing, ends up harming you and ends up being difficult.

Luisa’s intro [00:00:56]

Hi listeners, this is 80k After Hours. I’m Luisa Rodriguez, one of the hosts of the 80,000 Hours Podcast.

In today’s episode, I sat down with former-80k Advising Manager Alex Lawsen to talk about the mistakes he sees people make when trying to do good with their careers, and his advice on how to avoid them. This is advice I very much needed to hear seeing as I’ve made 7 of the 10 mistakes we talk about in the episode — and I’m guessing many regular listeners will find it resonates with them as well.

Since we recorded this conversation, Alex left 80k to join Open Philanthropy, where he’s now working on the AI Governance team… Which is why I’m extra-glad we got a chance to have this chat while he was still focused on giving great career advice.

Alex and I cover:

  • Not trying hard enough to fail
  • Why shorter AI timelines don’t necessarily you should try to work on AI right now
  • The downsides of constantly considering other potentially higher-impact career options
  • And much more

So without further ado, here’s Alex Lawsen.

The interview begins [00:02:03]

Luisa Rodriguez: Today, I’m here with Alex Lawsen, who manages all of the 80,000 Hours career advisors. We’re sitting down to talk about some of the biggest mistakes he sees people make, especially early on in their career, while trying to figure out how to have the biggest impact with their career. Thanks for doing this, Alex.

Alex Lawsen: Thanks for having me on. I’ve been really enjoying the couple of episodes I’ve heard from you so far, and I’m honoured to be one of them.

Luisa Rodriguez: Aww, thank you. That means a lot. OK, so today, we’re going to go through a long list of mistakes — so things like thinking you need to have a huge impact right away, or not trying hard enough to fail — and the plan is to kind of talk through those mistakes, what they look like in practice, why they happen, and how to avoid them. But before we get into it, can you say a bit more about how you came up with this list?

Alex Lawsen: I think there was a cluster of things that I noticed were unusually common in some people who shared some characteristics of the people that we advise. And once I noticed, I then realised there were other ones, and was thinking about it and decided, yeah, let’s see if we can talk about some of these things in a way which will maybe help more people than just the people we can officially speak to.

Luisa Rodriguez: Yeah, nice. In preparing for this chat, I framed a bunch of these patterns you’ve noticed as “mistakes” here. But I know that you’re a bit unsure about whether that’s the right framing. Is there anything you want to say about that before we talk anymore?

Alex Lawsen: I think “mistake” does feel like the most accurate framing. These are ideas people have, or patterns of thought or allergies they have, that I broadly think I want to try to correct. I think I was worried because often when I speak to people who have one of these patterns, I don’t think what’s going on is that they’re at fault. And I think that some people can quite naturally hear, “You’re making a mistake,” and kind of go, “Oh, wow. I suck.”

Luisa Rodriguez: Yeah. “I’m a bad person. I did this wrong.”

Alex Lawsen: Yeah. And I really don’t want that to be the thing. Maybe this could be framed as something like people are listening to advice that is not the most appropriate advice for them, or maybe they even need advice that’s kind of in the opposite direction to most people.

So yeah, maybe I could go for “adjustments to our advice” or “adjustments to advice people are hearing.” But, yeah, I think with some big caveat that’s like: It’s not your fault. There’s a bunch of things here that seem really hard. And I do actually think most of the things we’re going to talk here about are just mistakes.

Luisa Rodriguez: OK, yeah. That makes sense.

Taking 80,000 Hours’ rankings too seriously [00:04:41]

Luisa Rodriguez: Well, let’s dive in. So one thing you hear a lot is something like, “I should just do the thing that 80k says is best.” Can you say more about what that looks like?

Alex Lawsen: Yeah, I think there are a bunch of different ways this can come up. Maybe the most obvious one is just that, at any particular time, there’s going to be something really salient that it seems like lots of people should do. And maybe it’s because it’s at the top of the ranked list we have. That is one disadvantage of having ranked lists. So a very clean version of this mistake could just be people going, “The thing that’s at number one of the list of career profiles on the 80k website is technical AI safety research, so I guess I have to do technical AI safety research.”

So why might this be a mistake? The obvious thing is just that personal fit really matters. It matters in a bunch of ways. My guess is, during a lot of this conversation, I’m going to end up saying, “Did you know that personal fit is a big deal?” But specifically in the case of someone correctly realising that a thing is really important for more people to do, and then still making a mistake by thinking that they have to do it, what’s going on is something like it’s just a really bad fit for them. And I think there’s one way which is easy for people to internalise this, which is like, “I wouldn’t be very good at this, because I’m dumb or stupid” or something.

Luisa Rodriguez: Sounds familiar.

Alex Lawsen: And I always say that people are really different. People are good at different things. And not being able to do the thing that’s at the number one place, or the thing that your friend who’s read more of the website than you says is the best thing… Maybe there’s some sense where if you realise that the thing that’s going on there is you aren’t best placed to do that thing — maybe because you would hate it; maybe you’d be good at it, but you would just hate it and you would not manage to force yourself to do it for more than a couple of years — just realising, “That’s not the best thing for me to do. I should do something else.”

I have some guess that it’s easier to realise that maybe you should do something else if you realise that one of the things that you should be tracking is who is best placed to work on what. I can imagine some weird world where you’ve got this ranking of the best things for people to do, and then you just rank people by ability — but the distribution of ability is exactly the same across all the people: you just get like the best people and you put them in the best thing. Once that thing’s full up, you get the next best people, put them the next best thing. The world just doesn’t work like that. That’s maybe the thing that seems top of mind, is this: “I’ve heard one job is really important and I have to do that.”

And then a couple of ways this could go wrong that are really specific. One of them is try to do that despite it being a terrible fit, don’t manage to, and either get really sad and give up or just keep trying for ages and don’t go anywhere: try really hard anyway, despite it being a bad fit. That’s one kind of mistake you could make.

Luisa Rodriguez: Is there a pattern that you see often? Are there some kinds of roles? Maybe it is just AI technical safety research and you’re thinking something like there are loads of things to do in AI. If technical safety research isn’t the thing for you, if you notice yourself feeling like, I don’t feel very good at this, or it sounds weird and hard and boring, then consider doing something else in AI. There are loads of other things and those things could be really great even though they’re not literally the top of our 80,000 Hours career list.

Alex Lawsen: Yeah, maybe there’s a couple of levels at which this plays out. One thing I think does happen is people going like, “It’s technical or nothing.” Maybe there’s some societal thing here where people associate mathematical ability or scientific ability with generalised intelligence more strongly than they associate some other things that make you great at jobs.

The level of abstraction that people might make a mistake with there is going like, “Either I have to do AI technical safety or I suck.” I guess the mistake that happens there is, I do actually think people will, if they are a terrible fit for technical work, correctly rule themselves out of technical work, but then just conclude that they suck. And that seems bad to me, because there’s a bunch of other stuff that they could do that seems really great, and I don’t want people feeling like they suck.

Then there is a more specific thing that can happen. Let’s dig in within technical safety. So someone is pretty technical — maybe they’ve got a good programming background, maybe they’re good at maths — and they’re like, “I have to do the best thing.” And sure, AI technical safety is the best thing, but what’s the best thing in AI technical safety? Then someone might say you have to figure out the whole problem from first principles to have any idea of how to make progress.

And then it turns out that the technical thing that caused them to move in this direction was they’re just great at programming. They just write really good code really fast. Everyone else got sped up by GPT-4, but they didn’t get sped up by GPT-4 because they can write that fast anyway. You can tell by now that I’m not one of these people, because I’m pretty sure that’s not how it works.

But yeah, maybe they’re really good at something that isn’t sitting down and doing something close to philosophy, but they go, “The real value here is those philosopher kings, so I should try to be one of them, not a programmer.” And there, I think that how the mistake will look is, rather than not trying at all and just feeling like there’s nothing else that it’s worth even trying, and just not really bothering — or just trying something else but feeling really bad about it — they’ll actually just try to do the thing that maybe does seem best on some objective list, if everyone had equal ability, though I’m not sure how valuable that list is, rather than the thing that is way better for them.

Luisa Rodriguez: OK, so let’s say I come to you for advice. I’m a great programmer, but I want to do the most good. And the thing that supposedly does the most good is working very directly on AI technical safety by being some kind of philosopher that thinks about how to notice when an AI system is misaligned. What’s going through your head? What kinds of things would you say to me?

Alex Lawsen: Nice. I will answer this without making a claim about whether there is some ranking between doing an ML engineering role and something more like an abstract philosophy role. The thing I’d say to someone in that situation is: measuring the performance-adjusted difference between those two paths just seems like the wrong thing to be doing. I just find it vanishingly unlikely that someone’s going to be exactly as good at both. Actually, [or] if they’re a bit better at one than the other, but one’s way better than the other, maybe they should do the thing that’s way better than the other.

Luisa Rodriguez: Sure.

Alex Lawsen: But I’ll typically make an additional claim, which is: A, I expect the difference in ability to be pretty big, and B, that actually in lots of these kinds of roles — and this is actually especially true on the technical AI side — there’s just quite a big supply of talent that’s pretty good at maths and pretty good at programming. The bottleneck isn’t really people who might be able to do something pretty good at all; the bottleneck is something closer to really finding this specific thing great, like really being great at that specific thing. And I think people can hear a message like that and go, “Oh wow, I’m not a super genius, so I shouldn’t try.”

Luisa Rodriguez: So in this case where I’m your advisee, it’s something like I could be great at some programming role that relates to AI. And that might not feel like the most important thing broadly, but you’d argue that both kinds of roles are pretty important, even if we’re assuming here that one is maybe more important or valuable than the other. But just given that I could be great at one that’s still very important, means that being great at that is just probably better than being pretty good at the other.

Alex Lawsen: Yeah, I think the last thing is the key thing here.

Luisa Rodriguez: I do think a lot of people hear that, and I’ve heard that kind of thing and thought, “So wait, I can only help if I’m going to be great, if I’m going to be one of the best people at the certain thing?” I wonder if that’s going to trigger some people to want to back away, just because they’re not going to be able to be great at anything, or they worry they won’t.

Alex Lawsen: Yeah, I think that’s exactly one of the reactions I was most keen to talk about of all of the things that we might talk about. Because this is where I just really want to push hard against this idea that there are people who are great at stuff and there are people who are not great at stuff — and therefore if you are not great at a thing, you’re also going to be not great at everything else. This just seems straightforwardly false to me. There’s clearly some correlation in ability, especially when you’re in a kind of similar domain: like if someone’s an amazing physicist, my guess is they’re going to be a better theoretical chemist than an average person. But great at a role, or the role is a great fit for you, I think requires really exceptional interest and motivation in it as well.

And so if I’m talking to someone who seems like they could plausibly be great at several things, and the decision procedure they want to use is, “Which of the things is best? I’ll just do that one,” then I’m going to say probably which of the things you will contribute most in is way more determined by how good a match you are for that thing than which of those things is best in the average case. So what I’ll tell people to do is to try to test out how the thing’s going, look for some signals of it going well. And lots of those signals will be like, I’m enjoying it, I feel like I’m moving quickly, I feel like I’m learning.

Luisa Rodriguez: I feel like I’m crushing it.

Alex Lawsen: Yeah. And I think actually they do want to be looking for a signal that looks more like “I’m crushing it” than “This is going pretty well.” I think if someone is imposter-y enough or underconfident enough that they need reassurance about finding something that’s pretty great for you, I do want to say the signals that you’re crushing it should not only come from your internal narrative. Otherwise, you might just miss signals that you are crushing it.

Luisa Rodriguez: Yeah. You preempted my next worry, which is: What if you very rarely feel like you’re crushing it, and that’s just internal to you? It sounds like you’re encouraging people to have different kinds of evidence. Evidence that you’re crushing it might come from coworkers or managers, and you should put lots of weight on that if you’re the kind of person who’s basically never feeling like they’re crushing it.

Alex Lawsen: Yeah. If you’ve decided, “I’m a pretty technical person, I like thinking about AI, so I should try a bunch of stuff.” You’ve done this cool thing of going, “There’s a problem that I have some fit for, and I want to work on it.” And then I’ve said, “Find the thing that you’re best at within these three things.” And you try all of them, and it just doesn’t really feel like any of them are going that well, and you’re not feeling really excited to put a bunch of work into getting better at those things, which is part of the signal. If they’re not going super well, then I’m like, “Cool. OK, you’ve tested out that technical AI thing, and maybe try something else. You’ve just got some information about the variety of things that you can try in that space, and your fit is not as great as you were excited about, and you should try something else.” And this just seems fine. You are allowed, if you’re doing a test, the test has to have some outcome where it isn’t a success.

Not trying hard enough to fail [00:16:42]

Luisa Rodriguez: A possible outcome. Yeah. That brings us perfectly to another mistake that it sounds like you see a lot and is one that I’m super familiar with: not trying hard enough to fail. I knew exactly what you meant by this as soon as you said it to me in preparation for this conversation, because it’s a thing I think I’ve done dozens of times. Can you explain what this is for people for whom it’s not immediately clear?

Alex Lawsen: Yeah. What does trying hard enough to fail look like? It means that you are actually testing something. It means that you might get a signal that this thing that you are trying out is not the thing that you should be betting a significant chunk of your life on. Not trying hard enough to fail. What does this look like? It looks like leaving something in reserve — such that if stuff doesn’t go as well as you thought that it might, or thought that it should, for this to be the perfect career for you, that maybe the reason is just that you didn’t put in that little reserve, you didn’t try that little extra thing.

But I kind of feel like putting you on the spot here. The thing I do a lot isn’t explaining stuff in the abstract; it’s like responding to stuff people say.

Luisa Rodriguez: Sure. Yeah, let’s do me.

Alex Lawsen: So if you feel up for it, I promise the audience that I’m actually putting you on the spot here. Is there a time you think you didn’t try hard enough that you actually were able to fail?

Luisa Rodriguez: Yeah, definitely. I mean, I do feel like I’ve gotten the advice dozens of times that was like, “You’re pretty early career: test out this thing.” Early on, that thing was research. I applied for a job at 80,000 Hours, actually, and I had almost no research experience that was the kind of research that 80,000 Hours was doing. But I applied and I ended up doing a work trial with Rob. And I didn’t get the job at the end of the work trial, and it was really gutting. He then told me to try really hard at research, see if I could get another job — which I did; I got an offer from Rethink Priorities to do research — and he said, “Try really hard to nail it, to excel. And if you don’t, that’s really valuable information.”

And I don’t even know that we need to go into exactly what the outcome was. Researching at Rethink Priorities went pretty well, all things considered, but there was really no time while I was doing it that I was open to failing. I was pretending to be trying to fail, but mostly I was not wanting people to see early drafts so that they wouldn’t notice if I was doing bad research. And I was delaying publishing things so that I wouldn’t get comments from people saying, “This isn’t great.” And yeah, I mean, this has happened again and again.

Alex Lawsen: Yeah, I like this example. My guess is actually the thing I have in my head as most common is not quite this, but I like the example because I think I can point at a couple of different bits of it.

The first thing is I think Rob gave really good advice. And one way — which links to what we were talking about earlier, in terms of trying to find something that’s really excellent and excellent for you — that I think Rob’s advice was good, is that, ideally, what you want to do is you want to try really hard, and then have the bar for success be, “I’m nailing this, and I’m enjoying nailing this.” And if you’re trying really hard to make sure it’s going at least fine, and everyone else is seeing it going at least fine, then it’s going to be really hard for you to realise that there might be something that you could be even better at.

I think we’ve both talked a lot about our own mental health and having inflexibly high standards. And I think this is a case where actually you should have achievable standards — but higher standards than you might naively think — for saying, “This is the bet I want to make.”

Luisa Rodriguez: Right. And that’s because if you’re doing these kinds of tests, presumably there are a couple of different things that’s still plausible you’d be great at. And if it’s like one of the first two you’re trying, and you’re learning that you’re fine, it’s not great to settle for that option, if you end up choosing it and not trying something else, and learning that you would have been great at that something else.

Alex Lawsen: Yeah, exactly. I think maybe a case that I’m pretty sure I mentioned in the only other interview I’ve done on the 80k podcast was about how I approached exams when I was a teenager, and actually in university as well, which was really not preparing as much as my classmates would. And there were some other things going on that affected this, but certainly my internal narrative was like, “If I do well, having not tried that hard, then there’s nothing ruling out the possibility that I could have done astonishingly well if I had tried really hard.”

And that internal narrative is exactly the problem. The thing you want to do is rule out the possibility that this is the perfect match for you, so that you can consider whether something else is. And if you keep yourself in this limbo of, like, “Maybe this would be amazing if I tried harder at it…” Yeah. There’s something about, like, failing is really scary.

Luisa Rodriguez: It’s horrible. It’s really hard, it’s painful. It makes you feel bad about yourself.

Alex Lawsen: Yeah. And then it can also just free you up to try stuff that’s just much better for you and for everyone else.

Luisa Rodriguez: Right, yeah. Like, “Great, I have one fewer option for things. There are like a billion things I could do with my career. I don’t have to worry about that one.” Which is probably just good if you’re not feeling like you’re crushing it and enjoying it.

Alex Lawsen: Yeah. I do want to point back to the Rethink example for a second. There was a sense in which it sounds like you were doing this a lot at the task-by-task level of you’re writing some report and you don’t want people to give feedback on an early draft. And look, unless you’re terrible at research — and I just refuse to believe that’s the case, given that I’ve read some of your research — then if you put a tonne of effort in and don’t let anyone see early progress, you will reliably produce something that is at least fine, that is at least enough that no one is going to say, “This is a catastrophe. Just don’t ever bother again.”

And there’s some sense in which this is sort of safe in the near term, that you’re not going to hear, “Oh, wow. Try something else.” But actually, if you should try something else — because, just as a random hypothesis, it turns out you might be really good at podcasting — never getting that signal, never getting the nudge to try a different thing, ends up harming you and ends up being difficult.

Luisa Rodriguez: Yeah. I feel like this can look like a million different things. Are there other things that you’ve seen as an advisor, or other ways this can look?

Alex Lawsen: Yeah, I think there’s a few. Often these will come up in early-career people who are trying to test out a bunch of different things. Let’s say you think you might want to be able to do some kind of writing-based role. Maybe you want to be a journalist or a public intellectual, or even just like a successful blogger. And so you write a bunch of drafts and you never really share them, or you post like one every six months. It’s on the EA Forum, but not under your own name.

Luisa Rodriguez: Classic.

Alex Lawsen: And I think what’s going wrong here is just probably you’re not going to get a tonne of feedback either way, and you’re probably not going to get that much better at writing. Writing careers are hard. Forget impact: just from a personal perspective, you’re really going to have to have a bunch of stuff go well — not just in terms of talent for clarity or excision of ideas or new generative ideas, but even just how fast you can put stuff out. And yeah, if you’re just writing a little bit sometimes and then not putting it out there, you can do this for two years and still have no idea how good your writing is.

Luisa Rodriguez: Right. You get this ambiguous result. And then what do you do with an ambiguous result? You’ve got the nice feeling of like, maybe I could be crushing it, and I don’t have to stare in the face the fact that I’m unambiguously not crushing it. But yeah, you don’t know if that’s the thing you should be doing because it’s going fine, but not exceptionally well. Are there others?

Alex Lawsen: Maybe I should just say there’s a whole cluster of things that are like a thing you’re doing part time, but not enough of the time that you’re making progress. But I think a clear example here is upskilling without some sense of where you’re trying to get to and without pushing that hard to improve that quickly. I think this can be in a bunch of domains.

And I do want to distinguish this from learning more about a thing before you dive in: I think you should totally learn more about things before you dive in. But the signal you’re trying to get, if you’re reading for a couple of hours a week about something, is like, “Am I interested in reading more? Am I trying to find out more about this? And then if you decide you are excited to find out more, you are excited to learn more, continuing to do a couple of hours reading a week in your spare time: if we’re trying to see whether you should switch careers, you’re just going to still be in the same position six months later.

And yes, there’s some amount of like, how much time can I dedicate? And if you have a busy job, this is a hard question. You certainly want to have got some signal of excitement. But I’ve sometimes suggested to people that I would rather they booked off a couple of days or booked a weekend and just tried to sprint — for the weekend, directly at the thing, and saw how much progress they could make in that fixed time — rather than reading a little bit, but without any particular plan or structure and just goes on and on for ages. But so few people have talked about how much progress you’re supposed to make in an hour and a half a week that it’s totally possible you’re making really fast progress for an hour and a half a week.

Luisa Rodriguez: Sure.

Alex Lawsen: And that’s just like, you never get any signal. Maybe I can try and collect together: How do you know you’re trying hard enough at a test? I think it’s just that you can describe a situation where you conclude that the thing you’re trying is not working.

Luisa Rodriguez: Yeah. I feel like having success and failure criteria that are objective enough to measure yourself against seems really key here.

Alex Lawsen: Yeah. Here’s a recent example that could be kind of cool. There was a post on the EA Forum fairly recently about ways to test fit for biosecurity careers. I think it was aimed at fairly early-career people, and it was essentially just a kind of smallish research project: Can I produce a report? And it doesn’t have to be original research, but can I produce a summary of some part of this field in 10 hours — five hours a day across the weekend? And then I’ll send that to some people and see whether they think it’s nonsense. But even before I send it to anyone, you’ve got some signal on: How was it to do that? How would I feel if my week was doing that?

Luisa Rodriguez: Right. Did you enjoy it? Did you get interested in the thing? Was it a total slog and you felt like you could barely finish it?

Alex Lawsen: Yeah. And I think often this can just come quite quickly. And I do want to say there’s some amount of maybe it is worth persevering past the initial, “Ugh, this is rough.” It’s a new thing. It depends a bit on the people, but I can imagine something going wrong where someone’s like, “Alex said that all the stars have to line up. And I’ve never done a timed exercise that lasted more than an hour before. And I’m 90 minutes into this thing and my brain seems pretty tired, so I guess all the stars aren’t lining up.” Maybe if you’ve decided to do this sprint-through-a-weekend thing, maybe try the whole weekend and then reflect back afterwards: “How was this in total? If I look back at what I write, am I happy with it?”

Luisa Rodriguez: Yeah, there are all sorts of false negatives I’m sure you could get. Like you’re trying to do an independent research project to see if you could be good at research, but in fact you’d actually be doing research under a manager, so it is kind of fine if you find it hard to stay on task and motivated while working independently. So it is important to keep those things in mind and make sure you don’t rule something out when you’ve set up a test that isn’t really testing the thing. But it does make the importance of setting up the right test seem kind of key: make it so that it will be a reliable source of information about the actual thing that you are trying out.

Alex Lawsen: Yeah, I think that sounds right. I do actually endorse planning the test that you’re going to do, rather than just saying that you’re going to be learning more about it over time. Maybe get some feedback on what success should look like for the thing, and do some thinking yourself about, “Is a weekend sprint through a thing something that I would enjoy in general for anything?” I’m not wedded to this particular thing being the thing you should do. But I think it seems useful to contrast the other thing, which is, “I’ll do some research in my spare time when I have time, and maybe if that research goes really well, then I should switch into doing more of this.” And I’m like, if it’s in your spare time with no concrete plan, and you don’t even really know what the research is — it’s just “do some research” — then my guess is you’re never going to have any idea, based on that at least, how well this is going.

Luisa Rodriguez: Whether you’re great or terrible.

Alex Lawsen: Yeah, I think there’s maybe one other class of things here: One way you can get a signal on how well set up you are for a thing is applying to some jobs for that thing. This is something that comes up a bunch when I speak to people. They’ll say something like, “I’ve actually got really good evidence that I’m not going to make it in this field, because I’ve applied to a bunch of things and I’ve been rejected.” And first I’ll say something like, “Oh man, that sucks.” I don’t have to put on the empathy here; I am terrified of rejection in a bunch of contexts, and so I get that it sucks. But I will then ask, “Are you up for saying how many? And how far along the process you got, and what happened?”

And then sometimes it’s like, “I applied for these four hyper-competitive roles that I do think I was qualified for” — they won’t say “hyper-competitive,” I will fill in hyper-competitive — “and I got to the final round of two of them, and then I got to the one before the final round of the other two, and I was rejected from all of them. So clearly I can’t hack it.” And I’m like, it sounds like you weren’t applying ambitiously enough: if you’re getting past the first round of every application, it certainly sounds like you haven’t applied to enough things.

Luisa Rodriguez: Doesn’t sound like you’ve failed. Sounds like that could just be very consistent with applying for a bunch of jobs and eventually getting one.

Alex Lawsen: Yeah. And to be clear, I think there’s enough noise in especially the early stages of job application processes, that if someone’s applied for five things and been rejected at the first stage for all of them, I’m like, I’d probably have someone have a look at your CV and check you haven’t got a spelling mistake in it or whatever — check you haven’t been ruled out for stupid reasons — but probably just apply a bunch more times before you start updating on trying to get some more experience first.

Luisa Rodriguez: Right, so that’s one thing that can happen. But it’s also possible that you apply for 20 jobs that you thought you had a decent shot at getting, and you don’t get very far in any of them. And in that case, I guess both again, that sucks, and also, you’ve done a great job at actually trying to fail — and that might be a reasonable time to call it on that particular type of role.

Alex Lawsen: Yeah, I think that sounds right to me. Or certainly I think call it on trying to move in right now.

Luisa Rodriguez: Sure.

Alex Lawsen: And depending on the field and a bunch of other factors. This is why we have a one-on-one service: maybe 20 is actually just vanishingly low, or maybe 20 is probably you could have started calling it at 10. But yeah, I think this is a terrifying thing to do. I think there have been a few good pieces of advice, including on this podcast, about how to make this less painful. I also want to shout out someone I follow on Twitter who has these little getting rejected punchcards, which I think are kind of cool. You get ice cream if you get a certain number of rejections.

Luisa Rodriguez: I love that. No way. That’s great.

Alex Lawsen: Yeah, we’ll put a link into her tweet. I think there’s like an Etsy account that got set up where the money goes to GiveDirectly.

Luisa Rodriguez: That’s amazing. I actually do want to pause and talk for a minute about how to make this a bit easier, because it seems like it’s one thing to sit here and talk about why everyone should be trying harder to fail. I’ve personally basically never succeeded at this. I think I’ve gotten better at it. It’s just always stayed scary and hard and painful to me. I think a couple of things have made it better for me, and maybe there are some things that have worked for you that might be just worth mentioning.

So some things that worked for me were telling people I trust I was going to try really hard at a thing and might fail, and getting some reassurance that that seemed like a good thing to do, and also that they would understand and respect and love me even if I did fail. So when trying out this podcast host role, I told my partner how terrified I was of failing at it, and told him there was some chance it wasn’t going to work out, and that I’d find that very awkward and weird and embarrassing. But I needed him to tell me it was OK, and if we learned that, he’d love me and think it was just great that I tried.

And I found that just really reassuring and safe: in the worst case, I was going to have a partner that loved me and things were going to go on just fine. Yeah, I’m sure there are others. I’m trying to think of some off the top of my head. I don’t know if you have any others coming to mind.

Alex Lawsen: I really like that story. I think that’s really nice, and I find that’s really hard too. And maybe one thing to say is just: It’s hard. At least for some people, it is just never going to be that easy. Maybe I want to say something like if the reason you notice that you aren’t doing this thing of trying hard enough to fail is that you don’t feel like you’ve got a Plan Z, or you don’t feel secure in how stuff’s going, then my guess is to fix that before trying to do this really ambitious thing.

Luisa Rodriguez: Totally. Yes, that makes a tonne of sense. I was lucky enough with the podcast that even if the hosting thing went terribly, there was probably editing stuff I could do or help Rob research things for his interviews, and that made it a bunch safer. So yeah, that does seem really important.

I guess kind of piggybacking off of the ice cream thing, which I really like: Probably there are ways to set yourself up to get rewards for failing. One that I’ve tried a little bit is, again, told people I know I was going to try to fail, and then asked them to really praise me if I fail. I think there are times when I was applying for jobs that I did get friends to tell me like, “Congrats, you applied and got rejected!” Yay, round of applause. And in fact that made it better.

Alex Lawsen: Yeah, I think that’s certainly a cool thing to try out. And I remember one of our colleagues talked about something similar a long time ago on the podcast when it came to philosophy academia positions and people having spreadsheets and competing for numbers of rejections. I think different things work for different people. And yeah, I can imagine that there are some people who will be extremely reassured by this “my friends are going to say that I am great anyway” thing — and I think other people are honestly just going to really hate that.

Maybe for those people, one thing I’ve sometimes said is something like: If you have a success criterion for your application strategy, then you can be checking whether your strategy was correct. I think my success criterion for an application strategy is that there’s some part of it that’s numbers and then there’s some part of it that’s like, what’s the range of ambitiousness or the range of roles you’re applying for? And if I’m getting to different stages in different processes multiple times each, then probably I have a good spread.

And then maybe it depends a little bit on what your situation is. I think if you’re employed in a job that seems pretty great for you and you’re trying to do the occasionally ambitious thing, this sort of advice doesn’t apply as much. But in general, if you’re looking for work or looking to make a switch, trying to have some kind of pyramid where mostly you’re getting rejected at the first stage, but a decent chunk of the time — but less than half the time — you’re getting through to the next one. And then I don’t know how many stages there are, it depends on the job, but something like less than half of the time you’re getting one stage further. And that just requires you’re going to have to make a bunch of apps in order to get the job. But I think you are going to have to make a bunch of apps to get the job, speaking in very broad terms.

And so maybe there’s this thing that’s like, if I’m tracking this pyramid, then it looks like my strategy is working because this bottom tier is staying at the tier where it needs to be. Maybe something like that could be helpful for some people.

Luisa Rodriguez: That makes sense.

Alex Lawsen: There’s maybe one more thing, and I think this is the thing that I have found personally most helpful. It’s coming back to this thing I’ve touched on a bunch of places, where different people are good at different things. Let’s say I’m applying for something that has this difficult work test. We talked about the sprint: multiple hours doing the same thing, like writing some big document. I’ve just not done multiple hours of writing a single document in timed conditions. That’s not a thing I’ve practiced in all of my life. And yeah, I think there’s something about noticing that me performing badly at that is just not a huge signal about how good I am as a person, or how good I am even at other things. In my case, that sort of thing took someone else pointing it out to me, and not just once.

But I think there’s something about noticing what an application process is actually testing: both noticing that there’s some measurement error in it, there’s some randomness, there’s some luck, but also noticing that you’re being tested on a thing, and probably that thing is not the only quality that you think is relevant to a person’s worth.

Luisa Rodriguez: That sounds totally right.

Feeling like you need to optimise for having the most impact now [00:40:39]

Luisa Rodriguez: Well, there’s probably loads more we could say about that one, but let’s move to another one. So another common mistake that you hear is people feeling like they need to optimise for having the most impact they can right now. Why do you think that is a mistake?

Alex Lawsen: Yeah. Why is this a mistake? Let’s say that with some part of your time or effort or whatever, you want to do the most good. The thing that I claim most people care about, when they are trying to do this, actually is the most good across your life, or across the time that you’re working. Let’s say I’m a millionaire. I won the lottery. I won a million pounds, and I want to give that away. And I’m going to give away £1,000 a year for the next 1,000 years, because I just set myself up badly. I claim the thing I should be thinking about is how do I do the best with that million? — not how do I do the best with the first thousand — even at the cost of losing money. What this might look like in the money case is that maybe I should hire some researchers to help me work out what to do with the rest of the money.

And I think the key consideration that I end up highlighting to people who I think are trying to do the best thing right now is something like: It might be that setting yourself up well to do more good later looks like not directly having as much of an impact right now. Because probably learning is pretty good if you want to have an impact later. Probably getting some signalling experience for a lot of careers, maybe doing a prestigious internship, maybe getting paid a lot of money: these things just often pay off later, and often trade off against doing the very most good with, like, the summer internship you’re doing this year, or in the first two years of your job just after you’ve graduated.

Luisa Rodriguez: Yeah. That makes tonnes of sense. I really like the analogy of donating money over time, and not optimising for the first thousand. I feel like I did this, still do this. It just really resonates with me as a mistake that’s easy to make. In my case, it feels like it was a combination of wanting to feel good about feeling and thinking that I am having a positive impact, hopefully, on the world; wanting to feel like I’m living by my values of doing the most important thing, and not a thing that seems pretty good.

And then also just totally something about wanting my peers to perceive me as doing good and living by my values. So wanting to work somewhere that’s, like, legibly high impact, and not just somewhere that’s giving me career capital, but doesn’t seem as obviously good to them. And I think I’ve had this particularly because being part of the effective altruism community, there are lots of people who all share some overlapping views on what’s allowed, what’s high status — and I just felt super hyperaware of that, and wanted the approval, I guess, from people around me, of doing the good thing now. It’s just a lot of things pushing in that direction. How do we help people push in the other direction?

Alex Lawsen: My guess is at least some of them are going to be helped by hearing you just describe that. You said a bunch of stuff there, and I want to pick up on a few bits of it.

I certainly think that part of what’s going on sometimes here is this peer approval thing. Maybe, actually, it’s worth saying something like: If you’re a member of this community, or adjacent to it, and you’re talking to lots of other people in a similar position, kind of like making it clear that you won’t judge your friends for taking things that seem pretty normal career capital building instead of the, you know, weird EA internship that seems really cool because it’s got like a four-letter acronym running it. Probably at least two of the letters are “I” because EAs are not good at naming stuff. Or maybe there’s a “Centre” in there.

Luisa Rodriguez: Yep. It’s definitely got a “Centre” in there.

Alex Lawsen: But yeah, I think part of it can be pressure. The main thing I do to push back on this in calls is just hammering the thing: What are you trying to achieve? You’re trying to achieve the most good — not the most good this year in exchange for less good later. I do think that part of the difficulty is that maybe people are really worried about suspiciously convenient outcomes. Where you like take the high-status job at the quant trading firm, where you’re going to get to play with fun maths puzzles every day.

Luisa Rodriguez: And earn a bunch of money.

Alex Lawsen: And earn a bunch of money, and be seen as super duper smart by your normal peers who don’t have any idea what this altruism stuff is. That’s a real draw. And I think if that’s a draw that you don’t endorse, because you really care about helping as many people as you can, then it’s easy to not realise that there are “helping as many people as you can”-style reasons for doing that thing anyway.

Luisa Rodriguez: Totally.

Alex Lawsen: I actually hadn’t realised, but I basically did something like this, following the reasons that I didn’t really endorse when I finished my degree. So I knew I wanted to go into teaching, and one of the options was to do a programme called Teach First, which is similar to Teach for America. You get a couple of weeks training, and then get thrown into a school which has a bunch of difficult aspects to working there. It’s usually in an area where lots of the kids have a bunch of stuff going on at home; often lots of the parents need a bunch of support themselves, which is difficult to provide as a teacher, especially a new teacher. And if you survive for a couple of years, approximately, you get a teaching qualification.

And I could also have gone into teaching via a normal route. And I essentially picked Teach First to show, in particular, one of my parents that I wasn’t taking a cop-out option by picking teaching, and instead was like, “No, look: I’m doing the really hard thing.”

What’s interesting is I think Teach First does actually have much better career capital value than going into teaching via a more traditional route. I don’t actually think I ended up using it, but I can imagine, if I were facing the decision now, it being sensible to pick the option I did end up picking. Not because I wanted to show people I was capable of doing the hard thing, but just for much more mundane [reasons], like a bunch of top consultancies recruit Teach Firsters once they finish their two years of training, and you can move into management direction pretty quickly if you want to go down that path.

Luisa Rodriguez: Right. Yeah, that feels really familiar to me. Like, well, I have a selfish or an unendorsed reason to consider a certain career path. So I should rule that career path out, and ignore the selfless and sensible reasons to consider that career path.

Alex Lawsen: I think that’s a really good summary of maybe the most common mental move that happens here. Maybe not explicitly, but I think there’s often something like that going on.

I also notice that I think this is often pretty strongly tied to the first thing we were talking about is. You know, there’s something about having something with an external stamp on it, something that other people will approve of. And then there’s something that might end up turning out better in the long run, but looks more like career exploration, looks more like building skills, looks more like forging your own path and thinking for yourself. And it’s just really hard to choose that option — especially if that option would actually be better personally for you — because if you choose it, you’re going to be worried that you’re choosing it for the wrong reasons.

Luisa Rodriguez: Yeah. That makes tonnes of sense.

Alex Lawsen: I wonder if it might be worth mentioning at least one specific example of a mistake that I see in this particular direction. One that I’ve seen a bunch of times is someone who’s at university is pretty well set up for a technical research career afterwards on one of the world’s most pressing problems. So having said that not everyone has to work in AI safety, let’s use an AI example, because I think that actually, most of the people I’m thinking of do meet this description.

Let’s say someone is doing extremely well at a CS or a maths or combined degree, like really enjoying it. Maybe they’ve published a couple of papers in undergrad, got some cool research experience. And there is a sort of EA group at their college, which is kind of a bit of a nightmare to organise. And I’ve spoken to people in this situation, where they’re like “But right now, I’m just solving problems in my classes. And, yeah, I did this research assistantship this summer that got a paper out, but it wasn’t on anything important or relevant to safety. It was just on this robotics thing because that’s what my professor was excited about. I want to be able to continue the research, but I don’t have time to do all of these things because so much of my energy is taken up with running this club.”

And I think in this case, I’m just like, if you don’t like it and it’s not fun, why are you still doing this? If you can find someone else that can run it, great. I’m really excited about some of the student groups that have been set up by people who are really keen to do this. I think there’s been massive impacts from some of them getting lots of people involved, getting lots of people excited, helping people just meet other people like them who want to talk about these ideas.

My guess is that is not primarily coming from people who feel a strong obligation to do this thing that they really dislike. Maybe just go and get really good at AI. And if that’s the thing you’re pointing at, running an intro fellowship to the five people that you managed to get to sign up is just not setting you up well for that. It’s probably way less important for the world than you actually trying a bit harder on your own learning and then being slightly more likely to make an important breakthrough later. There’s no reason to do this other than needing to do something right now and worrying about picking stuff that seems too convenient, because you’d actually enjoy it and find it fun.

Luisa Rodriguez: Yeah. So what does it look like to do this thing well, or to not make this mistake?

Alex Lawsen: I think in some cases, this is pretty easy. The shortest version of this is something like: Just follow normal career heuristics, or at least strongly consider just doing what the selfish version of you would do — and the earlier on you are in your career, the more strongly you should consider this. So maybe taking the framing from the one blog post I did manage to write: when you’re thinking about what to optimise for in your first couple of choices, I think optimise for learning: developing skills, and developing an understanding of yourself and which things you enjoy and which things that you might be good at.

These are two different kinds of learning. I think both of them seem really important. So one could be like, “Lots of the things I might want to eventually end up doing will involve being able to write code really well. Maybe I should take this prestigious software engineering internship this summer, because it turns out most people at Google are pretty good at writing code. So if I go and hang out with a bunch of them, I’ll get pretty good at writing code.” Yeah, that’s one kind of pretty normal career heuristic of just like, “This skill seems like an important skill that’s useful for a bunch of things. Can I pick the thing that lets me get really good at it?”

Another thing — which some people will choose to dismiss because I used to be a teacher, but I just actually stand behind — is: consider trying to do really well in school. There’s actually value in getting really good grades if you can. It opens doors for you later; it opens doors for you at the time if your professors notice that you’re the person that’s coming top in different classes. So not to say everyone should just be focusing on school all of the time, but that’s a normal heuristic that often is correlated with success later. And if you can, like, get a 4.0 average in the States, seems worth doing.

Then maybe a different aspect of this that’s still kind of learning related is learning about yourself. So I think one possible way you could make a mistake in terms of this “I should index on personal fit harder than I previously thought” thing is to say, “Well, I know I’m good at this one thing. So I guess all of my internships and all of my experiences and my first three jobs should be in this one thing.”

I don’t think that seems like a good idea. One thing that can be really valuable to gain information about early on is, like, how many different things do you enjoy? So trying a variety of different things even when you have a pretty good idea of which things seem best is the sort of thing that very much looks like helping you over the long run. And this is maximising expected long-run impact. It may just be that you try out three things, and you’re like, “Yep. I was right about the first thing. I should just push on that.” I claim that’s still useful, because there are some alternate worlds where you didn’t find out that the first thing was best, and that’s a really good piece of information to find out early.

Then maybe the last thing — so all of these are something like learning, so maybe “maximise normal career heuristics” is not the shortest; the shortest is like, “learn lots of stuff” — but I think paying attention to mentorship, and particularly mentorship from outside this community of people who are really concerned about doing a lot of good, seems really important. This might look like doing a research project, if you’re a current student, whether that’s an undergrad or a post grad. Which is like, if you got a chance to work with a world expert on a cool kind of research, and that research isn’t actively doing harm to the world, then probably just learning from that researcher about how to do research really well should be very strongly considered as a good option, even if it’s not precisely pointing at the thing that you want to do research on later.

Similarly in jobs: Do you know who’s going to be managing you? Does that manager seem like someone who is going to help you develop? I was preparing for a meeting actually this morning, and wrote the sentence, something like, “I think I have learned more from Michelle, who is my current manager, than anyone else in my professional career.”

Luisa Rodriguez: Wow.

Alex Lawsen: That’s a huge deal. If you can find a manager who is really good at management, or even just a really good fit for managing you, I find it totally plausible that that is worth taking a job that on several other aspects looks quite a bit worse — especially if you’re early on, and it’s one of your first few jobs.

Luisa Rodriguez: Right, yeah. Cool. I think there’s a lot of stuff there that’s kind of counterintuitive and that people will get a lot from hearing.

I guess a thing that is still, like, when I imagine myself hearing this advice and trying to apply it — especially if I imagine myself trying to apply it three to five years ago — I think it still just feels really costly to me to have to sacrifice the feeling of I’m living by my values, and I feel very proud of that, and it’s a thing I care about a lot.

I mean I just wish it weren’t the case, but it is the fact that I care loads that my peers also think I’m living by my values. Is there more from just, like, a sociological perspective that people can do? I really like this suggestion of talking to your friends about how accepting you’d be of them doing a PhD because that is what they thought was the best way to build a skill and set themselves up to do a high-impact career later. Is there anything else? I’m trying to think of what would have helped me. And maybe it’s like, practice a story that you think explains to yourself and to others why you’re doing what you’re doing in a way that makes you feel like you can kind of explicitly show how you’re living by your values, even if it’s not some obviously preapproved thing you’re doing now.

Alex Lawsen: Yeah. I think this is a really good thing to push on. Maybe the first thing that seems important is, if I was speaking to Luisa three years ago on an advising call, and she was saying stuff like, “I really want to live up to my values,” the main thing I’d push on is like, “Wait. Your values are about doing the best thing, not the best thing right now. That’s actually the thing you care about.” So I might want to push in the direction of, if I can convince you that this does actually seem best in expectation across your long-run career, then that is the thing that’s in accordance with your values.

And I don’t know whether I would make this next move with everyone, but I find it plausible I might even say something like, “You care about acceptance from people in this immediate social circle of do-gooders, not from the world at large. It’s actually pretty unclear to me that you’re taking the easy route out by doing the normie prestigious internship, rather than the weird self-sacrificing thing that pays you much less. Plausibly, among the people whose opinions you care about, this is the hard option. This is the option where you have to explain your choice, and you feel like it’s not something that has their EA-approved stamp on it.” And I’m like, “You want to live up to your values? You want to do the brave thing? That’s the brave thing.” Sure, the normal world won’t judge you for it. But maybe the normal world isn’t the world you’re paying attention to.

Luisa Rodriguez: Yeah. I do think it seems like it could have really helped for me to try to notice that me wanting my peers to think I was doing something high impact right off the bat was not me trying to do the most good. And that me actually trying to live by my values would have been me trying to care less about that and focus on the thing I really deep down believed was for the best.

Maybe I just should have thought about that really explicitly, reminded myself of it regularly, told some people I cared about, and told them that it was important to me that they understood and at least agreed that it seemed like a plausible way for me to do the most good with my career. And maybe that’s just like a brave, hard thing that people should try to do a bit more of.

Alex Lawsen: Yeah. I’m glad that feels like it resonates a bit. I think there’s another thing that might be worth trying here. Some of this is actually just like a useful career planning heuristic in general, which is you should actually have a plausible story for how the thing you are doing might help, and that plausible story should contain some kind of concrete success criteria. So this is a jargony thing that I’ve imported from teaching, but the way I tend to phrase it or, like, the useful way of generating it is something like, “I’ll know I’m doing this well because…” — and then something that you can actually measure.

And so yeah, I think “just do the totally normie thing” is rarely the thing I’m actually going to endorse. What’s going on here is first we’re saying that maybe living by your values doesn’t look like doing the thing that seems to produce the most, like, utils in the next three months. And then the second thing we’re saying is, “Now that you have this wider option space, look at the option space carefully and see which things seem best to do.” But you want to have some reason that they seem best to do.

And I think this might be a case where if someone is really suffering from this kind of moral-perfectionist-flavoured intuition where they want to roll out huge classes of things because they’d be convenient for them, probably you do want to tag in another person. Maybe it’s an advisor. Maybe it’s just a friend who knows you pretty well. But get someone to help you go, like, “I want to rule these out for personal reasons. Which of them should I rule back in for non-personal reasons?”

Feeling like you need to work directly on AI immediately [01:01:38]

Luisa Rodriguez: OK, so another version of this, or at least a thing that sounds related, is people think something like, “I have to have an impact in the AI space now because we might get transformative AI in the next few decades. So there’s no time to wait until I’ve tested out various options or until I’ve skilled up more. I need to try to help now.” To what extent is that a thing you see? Is that a good description of it? Are there other variations of it?

Alex Lawsen: Yeah. I think that if it’s in the next few decades, which I think is a reasonable view on timelines, then maybe I’m just like, sounds like you should still be doing the career capital thing. A few decades sounds like the length of a career. There is, however, a claim that’s more like, “AI is coming in the next few years, and any work after that point is going to be totally useless. So I have a few years to make an impact.” And that last thing does follow from the first two. If you know AI is coming in a few years and that all work afterwards is going to be totally useless, then you shouldn’t be planning a 15-year career where you spend most of the first 10 years trying to set yourself up well.

I think there’s a bunch of subtlety here, and “it’s a bit more complicated than that” is maybe my whole take — we should talk about some of the ways it’s more complicated — but this is a reason to focus on doing more sooner rather than later. And the more strongly you believe the very extreme versions of claims about timelines then the more strongly you should weight towards doing stuff now rather than later.

Luisa Rodriguez: Yeah. What is one of the ways it gets more complicated?

Alex Lawsen: Let’s say, someone might say “I have 10-year timelines.” Usually, I expect this “10-year timelines” to be standing in for the median — so they think it’s 50/50 that AGI will have come in the next 10 years. And maybe what that looks like is 15% in the next three years, and then 35% in the seven after that, then something like maybe another 25% in the seven after that, and then maybe the last 25% is concentrated over the next 100, which is just slowly decaying. So maybe you can imagine a bar chart type thing.

And I have a bunch of thoughts about this, but I guess the main thing here is that [predicting a single year] seems absurdly overconfident. To the point that I don’t think anyone has actually made claims like this, or at least, I don’t think anyone sensible enough to be even thinking, “How should my career be affected by timelines?” has made claims that are like, “It’s definitely coming in this year at this exact time.” I think as soon as you switch to thinking about how there’s some chance in this year, there’s some chance in that year, and there’s some chance in a bunch of years afterwards, then I think something that’s easy to miss — but really important for someone like me to point out — is that which year you expect to have the most impact in isn’t the same question as which year you think it’s most likely AI arrives.

Partly, this is just that you’re going to get better at stuff later. Maybe if we really carved it up brutally, let’s say like 50% chance in the next 10 years and then 50% chance in the next 20 years after that — and I’m pulling these numbers out of thin air; I’m not saying this is like a prediction I endorse — but if you think you’re going to be way more effective if you spend 10 years, like, I don’t know, doing a PhD at a top lab, then it might just be worth not having much impact in the 50% of worlds where it comes in the next 10 years, if you can pay for that impact with increased usefulness in the later worlds.

So that’s one way that things could be a bit more complicated than, “Wow, five years. I’ve gotta do stuff now.”

Luisa Rodriguez: Yeah. Can I just make sure I understand?

Alex Lawsen: Sure.

Luisa Rodriguez: So it’s something like, if I’m kind of early career, or if I’m just not yet well positioned to help work on AI in some way — but I think I could be in like 10 years — I should just accept the fact that, in expectation, I will do more good by training up to be super useful in the worlds where AGI doesn’t come for another 15 years. Because realistically, I’m just not going to be very helpful in the worlds where AGI comes in three. And so rather than have, like, a tiny impact in the world where it comes in three, or maybe even a negative one because it might be really easy to do harm, I should just kind of rule out the worlds in which I contribute to AI in the short term, and instead plan for the worlds where I might be able to help with AI in the medium term or the long term.

Alex Lawsen: Yeah. I think that sounds right as this extreme case. I think there’ll be a bunch of different considerations here, and there are all sorts of small adjustments on the margin. But yeah, that is like a extremised version of one of these, where if it takes like five years to get good enough to be useful, then it doesn’t matter that there’s some decent probability on AGI coming in less than five years: you’re not going to help in those worlds anyway. So I think that is one consideration that seems important here.

Another version of this actually — which I think is interesting, because it doesn’t actually relate to career capital in the normal way — is something like, it might turn out to be the case that like a seven-year research project is required to solve alignment. If that’s the case, we’ve just already lost in all of the worlds where AGI comes in the next seven years, even if we start today. That doesn’t mean we shouldn’t do the research projects. Even if we find out with a 90% probability it’s coming in the next seven years, we should just be like, “Well, what can you do? We’ve lost.” We should still be doing the seven-year project because that’s our only hope.

And I don’t think there are things that are like that. But I think most people’s careers have something of that flavour — where you’re going to learn stuff and you’re going to gain experience and you’re going to gain skills, and that shifts how impactful your work is to later timelines compared to just the naive thing of like, “When does it seem most likely?”

Luisa Rodriguez: Yeah. It actually made a huge difference for me to have someone basically explain this thing, where I was feeling really guilty about maybe there’s some tiny chance I can help with, I don’t know, maybe AI governance in some way. And maybe it’s a huge mistake for me not to be doing that, given that I think it’s plausible that AGI comes in the next five to 10 years. And then someone was like, “You’re just not going to help in that world.” Like, maybe the 10-year world or something. But the extreme version, it’s just not going to be me. It’s going to be, like, mid-career people who have already thought about these issues a tonne. And it’s great that we have them. It’d be nice if I was one. But if I’m not, I shouldn’t then go try to do something that is unhelpful, and just a waste of my time and plausibly bad just because I feel guilty for not being involved.

Are there any considerations pushing in the other direction?

Alex Lawsen: I think there’s a few. So one thing that seems worth saying, which is not exactly about betting on shorter timelines, but I think it’s something that is often relevant to a decision like “Should I do a PhD?” because of timelines. There are plenty of other good reasons not to do a PhD. And I think one case where it does seem worth just trying to do the best thing straight away, whether because of timelines or otherwise, is because that thing would actually set you up best for the long run as well.

We’ve talked a lot about what if the best way to set yourself up for the long run is something that involves gaining skills or gaining capital or gaining mentorship or whatever. But one really good way of setting yourself up for doing a thing is just, like, doing the thing.

Luisa Rodriguez: Totally. Yeah.

Alex Lawsen: And look, if you could get a job working on something that seems like one of the most important things right now, you probably should do that. If you’re sacrificing a bunch of career development potential or skill development, whatever. Or if you’re picking something that seems like it has impact right now, but isn’t the thing you’re most excited about in the long run, there are reasons to not. But if the thing you think is most important for you to do is research engineering at an organisation that’s contributing to alignment, and you have an internship offer or a job offer as a research engineer somewhere with an alignment team that you rate, you should just do that thing. Because that just also helps you in the worlds where you do gain a bunch of experience in that role, and develop a bunch of skills, and it’s not in the next couple of years.

Luisa Rodriguez: What does it look like when people are making this mistake? So it’s like they think they should work on AI now. They can’t get those internships. What do they end up doing instead?

Alex Lawsen: I think basically something that doesn’t actually seem very useful, and is potentially quite severely harming their ability to learn and develop or be employed later, but that has some chance of being helpful.

I think there are some independent research projects that can take this form. For some people, I think independent research is the best choice. I think it’s extremely cool as a community that we have the ability to fund researchers to work on things themselves. And in fact, more than one of my advisees has gone down this path, and I’ve been really excited about it as the best option. But actually, most of the people who I’ve been excited about doing this as an option have been pretty senior, and the independent research is mostly like, “I’ve already been doing a bunch of research in another field, and I want to switch over” — rather than, “I’ve just graduated, and I’m going to teach myself to do all of the things.” Even in that case, I can see it being good for some people, but you take a big hit in terms of often sitting on your own with very little feedback and mentorship. It’s just pretty hard.

Luisa Rodriguez: Yeah. It’s bad for your wellbeing.

Alex Lawsen: Also just not very legible if you then want to go and get a job later, including potentially a job at some major lab. And then especially if you’re turning down something that looks very good on some of the other aspects — like turning down a top engineering internship, that kind of thing — I’m like, hmm.

Luisa Rodriguez: Or turning down a PhD programme that’s an excellent fit with a great advisor who you think can do a lot.

Alex Lawsen: Yeah. I think that sounds like a good example. Though I will say that just so many people should turn down PhDs for reasons that are totally unrelated.

Luisa Rodriguez: Sure. Yeah.

Alex Lawsen: And then the even more extreme version is choosing to do something that you totally don’t actually expect to be your best option if you have this extra time, because it’s something that helps right now. I think the example I see most frequently is people who clearly should at least be seeing whether they can make it as researchers, who are doing community-building or organisational work while really not enjoying that.

To be clear, I think it’s possible to make the other case. Say, if you’re a great project manager — you’re just always on top of 500 things, and you never drop any balls however many you’re juggling. And you’re like, “Should I spend six months really thinking about the core of the problem? I’m like, “No! Just do the organising!” But yeah, I think a case where someone is considering doing something that is very clearly not their skill set, because they expect it to have impact slightly sooner, is one of the biggest red flags here.

Luisa Rodriguez: Yeah, OK. So someone who conceivably could contribute to research at an AI lab doing something like being an executive assistant for someone at an AI lab now, because maybe AGI comes in the next few years — especially when they happen to hate executive assistant work, or at least don’t like it, and could thrive and really enjoy the engineering work they might actually have wanted to do to contribute on AI risk. Is that kind of in that boat?

Alex Lawsen: Yeah. I think that is probably so clear-cut that I don’t expect people to make quite that mistake.

Luisa Rodriguez: I feel like people do make that mistake.

Alex Lawsen: Yeah. Maybe. But certainly, I want to say something like, if you’re listening and going, “Well, my case isn’t that bad,” I don’t think you’re totally off the hook.

Luisa Rodriguez: OK, so it can be even less clear-cut than that, and you’d still be nervous.

Alex Lawsen: Yeah. One thing that seems worth noting here is that I think timelines have to be really, really, really short for the argument about, “Wow, I have to switch to something that’s higher impact per unit of time to go through.

And to try and illustrate this, there’s a few factors that shift how impactful your work is going to be over time. There’s the neglectedness thing we mentioned, which shifts towards work being better sooner. There’s the you getting experience and skills thing, which shifts towards work being more useful later on. And there’s also the thing where the more we know about the shape AGI is going to be, the more useful work is going to be — at least according to some models of how alignment work proceeds, which are controversial — as most stuff in the alignment space is — but for what it’s worth, I buy these models. I think that most of the useful alignment work is going to be done very close to AGI.

If you believe me, or agree with me and the other people who hold that position on that, I think you should be prepared to consider options which look like, even if something is 10 years away, spending seven or eight of those years setting yourself up. And then like two or three years actually working on the thing. I think a PhD is kind of a classic example here. If you think that research scientist positions at major labs, that are actually going to care about the signalling that a PhD provides, are the place where you are going to be able to have the highest impact, it does seem pretty plausible to me that you should spend seven years doing a PhD — even if, to first order, none of the stuff in your PhD is directly contributing to solving the problem. To be clear, if you’re trying to set yourself up for this specific path, there’s a bunch of PhD research that does not seem like the best thing for you to be doing. My guess is it’s just not going to be that clear-cut — because most of the best programmes, especially given how people have reacted to recent news, will let you work on relevant stuff during one.

But I do think it seems worth considering spending, like, four-fifths of your career setting yourself up for the last two-fifths. Then I think as soon as you have some uncertainty about how long you have, I claim this pushes in the direction of, like, “Well, if I turn out to have more time already set up very well, I can pivot to try other things.”

Luisa Rodriguez: Cool. Yeah, that makes lots of sense.

Alex Lawsen: Yeah. I think there is something else — and this is actually in a response I wrote to Ben Todd on this post that we’ll link to — I think the neglectedness point is less clearly in favour of skewing short for an individual thinking about their career, compared to something like an organisation making grants or an organisation like ours raising awareness. Because if you are relatively senior — like, if you have bet on the “I’m going to gain a bunch of experience,” particularly like research taste development or management experience — then, if you’ve been in the field for a while and now there’s an influx of new people coming in, potentially, you can have a pretty big multiplier on your own skills by this now much-less-neglected field having a bunch of people who you can help steer to do useful work.

I think this is actually a pretty strong consideration in favour of some kinds of career capital building that does take longer. If you think you could be an unusually good manager or mentor or research lead, then the fact that we expect, on average, this field to become less neglected over time may well suggest that getting in now and setting yourself up well to steer that increased effort once it becomes available could be pretty exciting.

Luisa Rodriguez: Yeah.

Not taking a role because you think you’ll be replaceable [01:18:19]

Luisa Rodriguez: OK, another thing you think people get wrong is deciding not to apply for a role or not to take a role because they think they might be replaceable in some way. Can you say more about what this looks like?

Alex Lawsen: Yeah. I think replaceability feels like a big can of worms. And I actually think one of the best discussions about it that already exists is also a podcast. It’s our colleagues Ben and Arden talking about this on some episode that we’ll link because I can’t remember the title.

But their very high-level idea of replaceability is one that 80,000 Hours wrote about far before I think either of us joined, and I think it makes a lot of sense in the original context. This is something like: If you’re thinking about whether or not to become a doctor, there’s lots of applications for medical school, and there’s only a fixed number of places. And so choosing not to go doesn’t mean there’s one fewer doctor in the world; it means someone who is slightly less good at medical school entrance exams than you becomes a doctor instead of you. And maybe this does make some difference, but it probably doesn’t make as much of a difference as one whole additional doctor.

This is an interesting idea. There is some truth to it, and it’s very… “sticky” is the word I want to use. Like, it’s this thing where it’s like, “Oh, wow. Have you considered that?”

Luisa Rodriguez: Right. It feels kind of mind blowing. Like, if I don’t apply for that, someone else will. And then my impact in that role would have been approximately nothing, because the only impact I could have taken credit for is the tiny marginal improvement I would have brought to the role.

Alex Lawsen: Yeah. So why do I think this is a mistake to be tracking, especially for the people that I expect to be forming the majority of the audience for this — people who are really interested in doing the very best thing, listening to lots of our advice, maybe making some of the mistakes we’ve talked about in terms of, “I have to do an EA-approved job”?

I guess one thing is that, depending on what you believe the distribution of talent looks like, it might not even be the case that the difference between the person who gets hired and the person who doesn’t get hired is really small. There are some fields where the very best people are just miles better than the next best, which means that there are some ways of arranging people, some distributions of talent, where the more people applying, the bigger the expected difference is between the first and second candidates.

Which is super counterintuitive. Our colleague Rob put some Excel or Google Sheets models together about this that we can link to. I guess maybe more convincing than Rob’s playing with spreadsheets is that I’ve spoken to a bunch of hiring managers. I’ve actually been a hiring manager on a few rounds. And it is just the case that some jobs require a tonne of skill — and in particular, a tonne of different skills, where they all kind of combine. And especially if they start combining more multiplicatively, it can just be the case that a perfect fit is miles better than a good fit.

So that’s one aspect where maybe this whole replaceability thing just isn’t even that true. And maybe it is especially untrue if the stage at which the selection is happening is a pretty intense application process immediately before doing the job, which looks pretty different to taking some general mental ability and memorisation exams at age 18 to determine whether you go into medical school. I realise that most non-British countries don’t have undergraduate medicine, so it’s like 21, but you know what I mean.

I then think probably that the more important consideration — if you’re thinking about jobs that do have this 80k-approved stamp on them or this EA-approved stamp on them — is: who are the other candidates that are pretty close to the top along with you? Who are the people that you are replacing in the role if you take it? Well, they are people who were really interested in having an impact by working for the kind of organisation that you’re already excited about working for, and they were pretty good at it. You’ve freed them up to go and do something else. There’s this really weird thing where, if what you care about is good being done rather than you doing it, then someone else not working at the job that you ended up getting and therefore choosing to do something else? Probably that “something else” is going to be pretty great for the world.

Now, obviously, the jobs that they get, they’re replacing someone else, and so then what did that person do? And then there’s a bunch of effects that all kind of knock into each other, and it seems really hard and complicated. So how do I cut through this?

My best guess is that if you’re applying for a job where really caring about doing good is one of the things that is a part of that job… Or maybe even not really caring about doing good in the abstract, but really caring about a specific problem. So for alignment research, if you care about AI being aligned, that’s plausibly enough; you don’t have to be this impartial welfarist who’s bought into longtermism. It’s mostly just like, do I think this thing should try to do what we want? Once you condition on that, my best guess at the strategy you should use for deciding whether or not to apply is: “I should apply to the things that I think I would do the most good in. And then I should pick the one that seems like I would do the most good in it.”

And how do I justify this? It’s something like if I zoom out and look at the whole community, the problem stops being, “How do we calculate 500 different counterfactuals and assign credit?” It’s like, “What’s the most efficient way of allocating talent between problems?”

Probably the most efficient way is that all of the talent tries to work out what they would be best at that’s also really good for the world, and then tries to do that. And then the people who seem best at the thing, and that thing seems best for the world, they do that. And then we get something where the most exciting jobs get filled up by the people who are the best fit for them, and then it trickles down and we just get a bunch of good people doing good work. And it seems way, way harder to do some weird central planning thing, where people agree not to apply for some things because someone else might be better at it.

And then on the organisational side as well, it seems important to do some amount of, if you’re working a really impactful job already, it seems fine for me to say, “I think there’s another really important impactful job you could do.” And part of this is this whole talent allocation thing. But part of my intuition from here actually comes from our colleague, Michelle, who was doing a really impactful job. She was running the Global Priorities Institute before someone suggested she might want to try out advising. And if you think about the status of 80k advising and the status of the Global Priorities Institute at that time, I can imagine someone who was living in this, “We’ve gotta worry about replaceability. And we’ve gotta worry about not poaching the best people for other good things because that seems bad. We should keep people in good things,” as soon as you start galaxy-braining this, I think probably no one suggests to Michelle she might be really good as an advisor and really enjoy it. And I think that would be bad for the world. I think it’s worked out well. It certainly worked out well for me, given this is the same Michelle who I said has been the best manager I’d ever had earlier.

Luisa Rodriguez: Yep. It feels super compelling to me that we have a bunch of individuals who don’t really know what the talent pools are like, and maybe they’re not that calibrated about what they’re excellent at — in many cases, I can imagine, as an imposter syndrome-y person, people getting it wrong in the direction of just thinking that they’re going to be worse than other people who could be doing the job.

You have all these people in the talent pool with like meh information — they know a bit about themselves; they know a bit about the roles out there — but for them to be trying to make decisions about how replaceable they will be and what the net effects on the community are is just a kind of bizarre thing that doesn’t really make sense. The thing that makes more sense is just for actual hiring managers to have as much information as they can about the pool. Which means they should probably apply widely, and then hiring managers know entirely about the pool, and they’re in a much better position to be like, “Who is the best person for this job? I will make them an offer. If this is the best job for them, they’ll take it. And if it’s not, I’ll take the second-best person for the job.” And the best person who might be really good for something else can go do that. That’s just like a much more efficient way to solve the problem, or at least get closer to a solution.

Alex Lawsen: Yeah. I think that gloss sounds right. So at the individual level, basically what this looks like is, if you are applying for jobs where you expect most of the other top candidates to care about some similar set of things as you do, then probably don’t worry about who you’re replacing or who might replace you, and just make some applications. And then on the hiring manager side, I think it looks kind of similar: if you think this is a really important role to fill, probably try to pick the best person for the role and let them choose whether that seems better than whatever else they would do.

I think the one place where I think you can do something more thinking about counterfactual replaceability is if you’re in a position where you’re holding multiple offers for roles, and they seem pretty similar in terms of fit for you. Or at least both of them seem plausibly like a great fit for you. I do think it can be reasonable to, at that stage, talk to the hiring managers about what the rest of the candidate pool looks like. And I would endorse — certainly within a community of people where everyone’s kind of pointing at the same broad set of goals — I think it does seem reasonable even to say things like, “I’ve got an offer from you and also from this other organisation. I think that these roles seem like a similarly good fit for me. I think these roles seem like a similarly good fit for the world. Probably, there’s some extent to which you’d be pretty happy about that other role being filled by someone good. Could you give me a sense of, either how close I was to the rest of the sample, or how easy you would expect it to be to fill this role if I didn’t take it?”

And this relies on both organisations being equally transparent for it to work. But I think that might be the case. I think, at that level, it can make sense. And it might just be the case that one of the orgs is like, “Actually, we think you’re the best, but there are three people who seem pretty damn good, and they’re not that far behind you. And our hiring process is kind of noisy.” And the other one is like, “We’ve been hiring for nine months, and you’re the first person that’s passed the bar, and you’re miles above it.” In that case, yeah, go with the second org.

Luisa Rodriguez: Yeah. If they do seem totally equal, that makes sense.

Constantly considering other career options [01:29:24]

Luisa Rodriguez: I think we can leave that one there and talk about another mistake. Another one that I’m super familiar with is basically being on a reasonably high-impact path or in a high-impact role, but still constantly considering other career options that might be even more impactful, in a way that’s either kind of distracting or leads to too much switching, and never getting great at one set of skills.

I feel like for years I was doing this. I think I’d been working for six or seven years, and I’d never had a job for more than 12 months.

Alex Lawsen: Wow.

Luisa Rodriguez: Yeah. And that was basically, I’d be in a role that I thought was reasonably worth doing, seemed kind of valuable. And then some low self esteem-y thing creeped in, where I was like, “What I’m doing probably isn’t that impactful, because it’s me, and I surely couldn’t be doing anything that impactful.” If ever another opportunity was available and seemed like it might be high impact, and it made the difference between my role and that role seem artificially big to me. I was just like, “Ooh, a chance to actually have a big impact instead of do this bad thing I’m currently doing, which is not impactful because I’m bad.” It wasn’t quite that extreme, but I think that was a thing pushing in that direction.

I think there are probably other things going on too. There’s a thing that can happen where it feels interesting and exciting and enticing to potentially get to do an even more impactful thing. And so whenever something’s presented, you’re just like, “Ooh, new thing! Potential for a new, bigger impact.” And for years it just meant I ended up being in a role for a bit, and someone suggested I apply for another one. In some cases, I got those roles, and then I’d switch because of a bunch of these biases, and then spent very little time getting actually very good at one thing because I’ve done it for years or something.

I guess to start, I’d be most interested in why you think this is so costly.

Alex Lawsen: Yeah, I think there’s a difficult balance to strike here. It’s not the case that people should just pick one thing and then go, “Great, this is my thing.” And I think people make this mistake a lot too, especially if there’s some “profession” that you’re in. So I think being a teacher, being a lawyer, being a doctor feels like one of these things where, like, “This is the path I’m on. Now all of my options are things within this path.” So yeah, it is hard to strike a balance, because I think people realise this and then they go, “I want to do the best thing,” and then they’re always looking for something else.

Where does this go wrong? There are a couple of different ways. One way — and from the sound of it, this has happened to you before — is you keep switching and you don’t get that much practice. Skills are transferable, but not all of the skills. Or even just you don’t settle into your groove, you feel kind of like, “How much should I even be investing into this place? Probably I’m going to have to switch in a year when I find something better.” That’s one way this can go wrong, and I think emotionally this can be pretty hard.

Luisa Rodriguez: Yeah, that’s true. Part of it was like, “Ooh, shiny new potential for impact!” But part of it was very moral obligation-y: A potential role was on the table, it seemed like it might be really important, maybe someone was trying to convince me it was super important. And then I’d feel like, “Best practice in the rest of the world is not to leave a job after nine months, but we’re not in the rest of the world. I’m trying to have the biggest impact possible. So I shouldn’t be bound to these norms. I should really seriously consider this job if it actually seems more valuable” — and then I’d just get really shame-y if I didn’t consider it seriously. I’d be like, “I’m not biting the bullet and trying to do the most good. I’m just trying to live my life comfortably.”

That was a pattern for years, and did basically mean that every nine to 12 months I was getting really sad and feeling really bad about myself, and maybe I wasn’t trying to do the most good.

Alex Lawsen: Yeah, that sucks. And I speak to a lot of people who have this kind of dynamic. I think one thing that’s important to highlight here — and I do want to talk about a framing I like that I think protects against both failures — but I think in that position, failing to move also isn’t going to go well for you. If people are successfully making you feel guilty about not considering other options, one mistake or one difficult outcome of this is that you switch to do the thing that seems better, and then you switch to do the thing that seems better, and you never get that good at any of the things, because you switch to something else.

Another thing, and I see this actually quite a lot too, is people are doing something that seems to me objectively great, and they’re just feeling terrible about it and not putting that much time or energy into it, because they’re doing a bunch of job hunting on the side. Because what if it’s not the thing that’s objectively great? What if there’s something better? And applying for jobs takes time, it takes energy, it takes emotional resilience. And feeling like you constantly have to be putting some of your attention on whether there’s something better, even if you don’t then switch to it, still results in you paying pretty significant costs.

Luisa Rodriguez: Yeah, that sounds totally right. So what is the better decision-making process, if this is the kind of trap you fall into?

Alex Lawsen: I have a thing which has worked really well for me. I think it works well for lots of people, and probably won’t work well for everyone. I think I’ll just try to describe the thing, but then I’ve got a metaphor I’ve been playing with that maybe I want to try and describe as well. So the thing is, you stick to some policy, which is like: “I’m going to look at a bunch of things, I’m going to actually seriously consider my options. And then with all of the information I have, I’m going to make a decision. And I’m going to make the decision to do the thing that seems best for some fixed period of time. At the end of that fixed period of time, then I will consider other options.”

And I think for most people, you don’t have to make this a fully binding commitment, that you just instantly say no as soon as anyone says, “Have you considered…?” I think of it as a strong default. If someone says, “Hey, have you thought about doing this other thing?” I can just say, “The next time I’m looking at options is in a couple of years. If you want to pitch me on this really hard, I’m not going to rule out ever doing anything before that time, ever, but you should know that you’ve got a lot of work to do to convince me to even really think about this. Because my plan is to stay for this long.”

Luisa Rodriguez: Right. Strong default is I stay doing this for two years, because I said I would. I thought about it really hard. It seemed like one of the best options. And there are so many costs of constantly considering these other opportunities that I’m mostly not going to — again, unless you want to make the case that it’s really important that I do, and then maybe we could be convinced. And it sounds like you’re allowing for that.

Alex Lawsen: Yeah, I allow for that. And I can imagine it being good for some people, especially people who fall more on the axis of they will actually stay at the thing, but they’ll feel guilty about not considering the other options. I can imagine some of those people going, “No, I actually made a commitment that I don’t want to break. Maybe I made a bad decision in making that commitment, but I made it and I stick to my commitments.” I think that could work for some people. I get plenty of reassurance from the strong default thing.

Luisa Rodriguez: Nice. Yeah. You said you had a metaphor for this kind of thing. What is that metaphor?

Alex Lawsen: Yeah, maybe I’ll just try and describe it. The metaphor is trying to point out not the strategy; the strategy is pretty clear: gather some information, make a decision with that information, stick to that decision for some period of time that you’ve planned in advance, and then reevaluate, gather some more information, and then make a new decision.

The least bad metaphor that I’ve come up with is you’re rowing a boat on your own and you’re trying to get somewhere. You’ve got some map that you need to look at to see where you’re going. I imagine like a map and compass, but I think actually pulling out your phone and unlocking it and looking at Google Maps is maybe fine. And I speak as someone who has, like, I don’t know, ever been in a boat. But I gather that when you’re rowing, you’re facing back: you can’t see where you’re going. You’ve just got to sit there and pull both of the oars a bunch of times, and the boat goes forwards. And then I don’t know how you steer, probably pull harder with one side, something like that.

I think what’s nice about this is I can imagine — and if you’re listening to this, you’re just going to have to imagine me doing some really effective miming — I can imagine you sitting forwards in the boat and trying to hold the map with your left hand while it’s gripping the oar and trying to hold the compass with your right hand while it’s gripping: pushing them rather than pulling them while looking at where you’re going, so you’re always precisely on track. But my guess is you’re just going to go super slowly because that’s not how to row a boat.

Whereas you can imagine someone else, maybe someone that’s racing you, who is going to point the boat in pretty much the right direction. They’re not exactly sure it’s the right direction, and they might go a bit off course. And then they go, “Cool. I’m going to row hard for a minute, and then I’m going to stop and check if I’m pointing in the right direction, and then I’m going to row hard for another minute.”

Luisa Rodriguez: Yeah, I think that’s great.

Overthinking or over-optimising career choices [01:39:51]

Luisa Rodriguez: So on the flip side, another mistake people can make is spending too long working out if the path they’re on is the very best, rather than just doing the path and kind of doing well in it. Can you say more about what that looks like?

Alex Lawsen: Yeah, it’s funny because I think this has the same solution, actually: you spend a bunch of time trying to work out what the best thing is, given the information you have, and then you just try to do that thing for a bit. There’s maybe an additional reason that this works well in this case, which is that actually desk research, gathering information about what’s best, has diminishing returns pretty quickly. Sometimes to know if you’re good at a thing and whether you like the thing, you actually have to try to do the thing, rather than just read about what the thing is.

But yeah, it’s a similar dynamic, it’s a similar internal monologue often of something like moral perfectionism: “Is this really the best thing? Should I be doing the best thing? Am I taking the easy way out? What if there’s something better?” The difference with this example is people are noticing that they might have that feeling if they start a thing and are starting nothing instead.

Yeah, we have done some caveatting at various points about how “mistake” is not a perfect framing. And I do think that it is not the best way to actually be doing the best thing, to just sit and wait until you’re absolutely sure. Sometimes you’re just not going to be absolutely sure and you’ve just got to try something. But I think some people who end up in this position are actually tracking the possibility of the failure we just talked about, where they start doing a thing, and then they’re so worried about whether it’s the best that they’re just miserable, and they never find out if it is the best thing for them because they’re not putting all of their effort in, because they’ve got one foot out of the door because they think something else could be better.

I do want to say, “Well done for noticing that and wanting to avoid it.” And the solution isn’t just to start something while worrying about whether it’s the best thing. The solution is, well, you’re never going to have all of the information, so gather some amount of information that feels like it is a reasonable amount to gather. I claim that you should decide what is a reasonable amount of information to gather before you start gathering it — rather than gather information until you think you’ve worked it out and then just go with that for a bit — and you will learn more about yourself and about the thing you’re trying with that.

Maybe one thing is we implicitly were talking about a scale of a couple of years, because I think that’s about right for people who are where we are in our careers, and who are in danger of making a mistake of being in a thing and worrying about whether there’s something else.

I think if you’re relatively new and you don’t know what you might be good at, then it might be that you do an information-gathering thing which is closer to, “I’m going to spend a day thinking about this, and then I’m going to try finding out more about this thing for a week, and then I’ll reevaluate.” And the evaluation will either be like, “This is going pretty well. I should spend another three weeks on this,” or, “I don’t feel as great about this anymore. Maybe I’ll try something else for a week.”

The timescale of the information gathering and of actually just trying to do a thing should get longer the further on in your career and the more confident you are in the broad-strokes thing you’re trying. But I do think it’s very rare that the time you spend just purely gathering information is even like 20% as long as the time you spend actually doing the thing. I think probably you should spend much less time on just purely desk-based research, gathering information before you make a decision, and you should probably be mostly just trying to do a thing for a bit.

Luisa Rodriguez: That makes sense. So the fact that I once spent about a month deciding whether to do a role for 12 months is probably not the right way to have done it?

Alex Lawsen: I don’t know, I don’t think that’s insane. Especially if it depends how difficult the choice was and what the other options were. It seems better than spending six months wondering whether there are any roles you should apply to, and then applying to three. I think it seems pretty reasonable, if you’re in work, and there is one option that seems like it could possibly be better, and you haven’t already made some hard default commitment to “this is not the time I evaluate.” I think spending a month — while mostly doing your job, having some attention on whether you want to try something else — seems like about the right amount of time.

Luisa Rodriguez: So do you think the main thing going on here is something like moral perfectionism? It’s like, “This is such a high-stakes decision. I need to be absolutely sure I’ve picked the best thing for me, given I don’t know what’s important and my personal fit, and I need to nail all of those factors down before taking action with my career.”

Alex Lawsen: My guess is that’s one thing that could be causing this. I don’t have a great sense of the relative frequency. I think this is probably the main thing that using this pick-a-direction strategy is helping in my case. I think I have a lot of this, but I can imagine there are some other dynamics that lead to a similar place.

One other dynamic could be pressure from other people. And this is not quite the same flavour as moral perfectionism in some internal sense, but I think there are dynamics in the EA community where people are like, “But are you doing the best thing, though?” Luckily I think there are also dynamics where people just go like, “No, I have a policy about when I reevaluate this.” That’s just not as weird a thing to say as it might be in some other circles. So I guess that’s one other related but different dynamic that could be causing this kind of thing.

Luisa Rodriguez: Yeah. Is there anything else that could be going on there?

Alex Lawsen: Yeah, I think one other feature of this kind of problem is actually just that often the decisions are just ones that it’s really hard to be that sure of — like unusually hard.

Luisa Rodriguez: Right.

Alex Lawsen: And, yeah, all of these have some overlap with each other, but especially if you’re working on causes to do with catastrophic risk or potential future people, there’s just a lot of uncertainty: there’s a lot of moral uncertainty, there’s a lot of empirical uncertainty, and then there’s additional uncertainty about it being hard to notice what good feedback loops are.

And so maybe some of what’s going on isn’t that people are making a mistake in the abstract about how high standards of evidence need to be, or how good a person you need to be in terms of being sure you’re doing the best thing. But they’re more just not noticing that the position they’re in is one of unusually high uncertainty. So they’re saying, “I think ‘reasonably sure’ is the standard I should be meeting here, because for most decisions like this, most people would be reasonably sure.” And then what’s going on is that “reasonably sure” is actually just a way higher bar for the specific decision they’re facing than the average case for a person on this planet.

Luisa Rodriguez: Yeah, and maybe it’s worth saying that it can feel bad to be making a very big stakes-y decision — where we say on the website it’s “your most important decision” — and if you’re only a little bit sure that you’re making the right call, that does suck. It’s shitty that we’re making decisions under this much uncertainty, and you’ve got to be pretty brave to do it. And so it won’t necessarily feel easy or good, but props if you manage to make decisions under that much uncertainty without being overly perfectionist about it.

Alex Lawsen: Yeah, I agree. I think it is, at least in my books, an admirable thing to be able to do. I maybe also want to pitch for it being the right thing — or at least right under some sense of if you think expected value is a reasonable framework to use, in some sense, for this kind of decision — then I do actually want to say that I think having this kind of policy is actually the thing that seems best in expectation.

That’s actually a lot of what I am trying to do with these analogies I’ve been playing with. Like the rowing one: I just think that you are paying some cost in not pointing the boat in the right direction, and then you are getting a speed up and the boat’s moving faster. What does this translate to? You’re doing better at the thing you’re currently doing because you have more focus on it. Probably you’re also gaining more information about how well you’re doing at that thing because you’re focusing more on it.

So I’m just saying I think you don’t have the option of both knowing that you’re pointing in exactly the right direction, and rowing as hard as you can. Maybe that’s the bit that I just want to stick with: There is no choice you can make where you know everything that you could possibly know, and you’ve been trying as hard as you possibly can to just do the thing and not worry about gathering more information. The options you have are: spend more effort on the thing and less effort gathering information, or more effort gathering information and less effort on the thing.

Luisa Rodriguez: There are tradeoffs.

Alex Lawsen: Yeah. And I think actually noticing that there is a tradeoff here is maybe one way of making it feel a bit less stakes-y, or a bit less hard.

Luisa Rodriguez: Yeah, feel more like it is the right thing to do. It’s like, “I want to be morally good by making sure I’m doing the right thing,” and then noticing that you could also try to be really morally good by doing a thing that you think is most likely to be the best thing you can be doing. Doing it for a bit longer, trying to get better at it by doing it. Those are all worthy, good, important things too.

Being unwilling to think things through for yourself [01:49:31]

Luisa Rodriguez: Let’s talk about another one. Another mistake that’s familiar to me is the idea that I shouldn’t do my own cause prioritisation because I won’t make progress beyond what others have done, like people at Open Philanthropy or 80k. Why is that a mistake?

Alex Lawsen: I think to some extent, I think it is just fine. Certainly, I think the version where you say, “No, I can’t listen to anyone else who’s thought about this; I’ve got to reason through things for myself from first principles” feels like maybe an even bigger mistake.

Why did I still pick up on there being something here? I think that people will often do a version of this, which is like, “I actually have some intuitions or some initial thoughts, but clearly there’s not going to be any signal in those because these super-smart people in their ivory towers have figured everything out. So the reason I have those intuitions is because I’m bad or stupid, and I should just squash them and do the thing that someone else said was the best thing.”

I’m like, I don’t know, a bunch of stuff I’ve said elsewhere doesn’t sound like it’s going to go that well for you if you actually have intuitions pointing in a different direction. And the goal of you trying to think through these things for yourself is not to solve all of moral philosophy; it’s just to work out which things you want to point at, and whether you have done enough of an information gather that you can feel OK going, “Maybe I don’t know exactly what all of the fine details are, but something in this direction seems like it’s something that’s worth me trying hard to get.”

I think that’s maybe the level I’d want to leave it, even: just that you are allowed to think through stuff for yourself. If you notice that there’s some internal tension, spend some time thinking about it. Don’t automatically assume that the reason the internal tension is there is because someone else got the right answer and you just have to push yourself until you’ve got the right answer. But yeah, I don’t think that people should first solve all of moral philosophy and then try to do any job at all.

Luisa Rodriguez: Yeah, that just sounds super reasonable to me.

Ignoring conventional career wisdom [01:51:37]

Luisa Rodriguez: Let’s go ahead and move to another one. Another common mistake you’ve seen people make is not following conventional wisdom on careers as much as they should. I’m sure there are lots of examples of this, but is there one that comes to mind?

Alex Lawsen: Maybe the most important example is underemphasising the value of getting good mentorship. Partly this is a conventional piece of wisdom about careers. In particular there’s another level of the conventional wisdom aspect here, where I think lots of good mentorship is available outside of the specific focus that you might have for your career. This might be like, if you’re interested in being a researcher, doing a project with someone who is an academic, but not an academic who’s interested in exactly the technical field you want. And just there isn’t a tonne of mentorship in the very narrow fields that most people who are focusing really hard on impact are pointing at — partly because they’ve picked those for neglectedness, among other things.

You should be thinking about how fast you’re able to learn and improve, especially early on. And I claim — as a former teacher, to be fair, so there’s some bias here — that who is helping you learn, who can you point to as a good example, is a really big aspect of how good a job is for you.

Luisa Rodriguez: Nice. Have you had the experience of getting good mentorship? And how valuable was that for you?

Alex Lawsen: Yeah, I think I’ve experienced a variety of qualities of mentorship, and maybe this is why this is extremely salient, but I think some of the experiences I’ve had here, being managed by Michelle — especially as I’ve transitioned into management myself — have just allowed me to develop far faster than I would have done absent that. It’s just really useful to be able to make mistakes and feel psychologically safe enough to say, “Hey, I think I messed up this thing. Can we talk about what I maybe should have done instead?”

Maybe that’s a specific example of the kind of mentorship that can be really valuable, is someone who you trust enough and feel safe enough with to ask them about specific things you’ve messed up, and get feedback on how to fix them, rather than just try a different thing next time. And a large part of that is that you also want to rate them enough that you are going to listen to them when they say, “You should have done this.”

Luisa Rodriguez: Yeah. Why do you think so many people are failing to get this, given how valuable it is?

Alex Lawsen: I think part of it is that your ideal mentor is someone who has the same values as you, who knows you really well, and who is spending a tonne of time helping you improve at this thing, while also the thing you’re doing is something you’re already good at and is really good for the world. And that’s just a lot of things, and there’s not that many people like that for most people. And then in particular, if you’re in a small community of often pretty young people, it’s going to be unusually hard to find a mentor who is also part of that community.

So some of this is just there are not that many great mentors to go around. If you find one that ticks all of the boxes, you’re really lucky, and you should really be excited about that. But if you don’t find one who ticks the specific box of being a great mentor for you, then maybe I just want to say, upweight that consideration compared to several others — even “How much impact am I having right now, this second?” — especially if you are planning to be trying to do good for quite a while. I would encourage more people than are currently doing it, in my opinion, to sacrifice some immediate impact over the next few months or years for a really good learning opportunity where they can put themselves in a great position to do good work later.

Luisa Rodriguez: OK, so practically speaking, that might mean something like if two roles seem pretty great, and you’re choosing between them. One of them seems a bit better on impact, and the other one has exceptionally good mentorship. It’s not crazy to choose the one that’s plausibly worse on impact in order to get the huge boost from the mentorship.

Alex Lawsen: Yeah, I think that sounds right. And maybe another example of this would be something like, let’s stick with like an academic one: say you’re trying to find a supervisor for a research project, maybe for a graduate degree, and there’s someone who is known for just producing phenomenal students year after year. This could be that their students get into the top conferences; it could be that their students just go on to have really successful careers in other fields; it could just be that you’ve spoken to some of their current students and they’ve said, “She’s always there for me, and is really supportive with all of the stuff that I’ve struggled with, and knows so much about this field.”

And then there’s also someone who, maybe even it’s not that you know that they’re bad on any of these axes, but you just don’t know them that well. But they have an interest in clean air specifically for pandemic prevention, and that is the thing you’re really interested in.

I’m just like, ideally find out more information about both. Grad school is a big decision. Asking to speak to some current students of both seems like a good idea. But it seems possible that, if you’re going to be doing some relevant research in both, taking the mentor who is a bit less aligned with your ultimate values, but way better for you at actually helping you learn a bunch of useful skills in research, I imagine that’s the best trade more often than I perceive people to be taking it.

Luisa Rodriguez: Cool. Are there other examples of conventional career wisdom that people are ignoring or just not putting enough weight on?

Alex Lawsen: I think another one is just doing a normal job at a company or organisation that’s known for being run well.

I think there are tradeoffs. If you join some small scrappy startup with only three people, then it seems totally possible that you get to gain a bunch of responsibility really quickly, and that might be the best choice for some people. But sometimes just really established programmes that know how to grab smart graduates without much work experience, and turn them into extremely valuable employees really quickly, just are actually good at that. I don’t know, trying really hard to do a thing for many years probably makes you at least decent at that thing.

Maybe if I can think of some specific examples from different fields here, I notice that many of the people I am just incredibly impressed with on their ability to think about different quantitative technical problems have, at some point, worked at Jane Street.

Luisa Rodriguez: Interesting. OK.

Alex Lawsen: And to be clear, I think some of this is probably that, whoever Jane Street’s recruiters are, they’re doing a really good job of just selecting people who are super bright and talented. But my guess is that some of it is that there’s a bunch of competent people that work there, they know how to take a certain kind of bright, talented person and help them learn how to think about particular things in a particular way. And at least if that is the case, then that should upweight your potential decision to go in that direction.

I have a hypothesis which I think might explain some of this underrating these heuristics, and it’s something like “being too wary of suspicious convergence.” I’ve used that phrase because there’s an old blog post that I actually really like, called something like “Beware surprising and suspicious convergence.” We can link to that. You can read it. I’m not trashing the post; I think it’s a good post. But I think there is a version of this flag which is like, “This decision I’m about to make would be selfishly good for me. So I’ve got to be careful about picking that decision because I don’t want to be selfish; I want to be doing good for the world.”

And maybe all I need to say here is: Sometimes what’s good for you and what’s good for the world is the same thing. And don’t overadjust. Don’t go like, “Normal people just care about signalling, and just care about working for the biggest, most impressive company. And I care about what’s good for the world. So I guess I should weight signalling at zero and working for a big impressive company at zero.” And I’m like, “I was with you until you said zero.” You should still care about them at all.

Luisa Rodriguez: Yeah, that makes sense.

Alex Lawsen: I maybe want to reference another post that I’ve referenced before, actually in the blog post I wrote for the site: “Policy debates should not appear one-sided.” I think it’s actually pointing to a similar thing to the thing you described, which might be the thing in this suspicious convergence post, but it seems important to me. Which is like, if all of the reasons you have for doing a thing point towards doing the thing, then are you using motivated reasoning? That seems kind of likely. Often decisions are just hard, and different tradeoffs point in different directions.

And maybe what I’m saying in this “don’t totally ignore normal career heuristics or conventional career wisdom” is to say political debates might not be one-sided; they shouldn’t be perfectly one-sided. They also aren’t necessarily going to be perfectly balanced. Sometimes there are considerations pushing in one direction, and there are more of those than pushing in the other direction, and you shouldn’t go, “Oh no, both the good things for me and the good things for the world mostly say this thing. So maybe the good thing for the world has to be in the other thing.” Sometimes you don’t have a really hard tradeoff.

Luisa Rodriguez: Yeah. Sometimes you just get really lucky, and you get to do all the things that are good for you and the things that are good for the world. I guess it’s probably also the case that some types of these reasons will be more likely to trigger some suspicion. It’ll be like things that feel selfish to you. It probably varies for different people.

I guess that just seems traditional conventional wisdom feels like it’s often aimed at and for people who want a lot of money and power. And a lot of people trying to do good in the world think those motivations are at least not the most important ones, and maybe sometimes bad and corrupting. So it’s just really likely that you’re going to feel kind of allergic to things that get you those things. And that just seems understandable, coming from probably a nice good place that causes you to be more altruistic than an average person, but that can be a bias that might make you make worse decisions, or be suspicious of decisions that you shouldn’t be.

Alex Lawsen: Yeah, I think that seems possible. There’s this thing that’s like, “The people who care really a lot about this seem like kind of sus, so maybe I should care zero about it.”

I have some maybe slightly more general version of this that I think could be going on, and I think could be going on in a bunch of other places as well, which looks more like you’re with a group of people who have noticed that the world, on average, is making a mistake. And so you’re not going to make that mistake. And so here, maybe the world thinks status is too important, and you’re like, “Well, I’m not going to make the mistake the world makes.” And I don’t know how true this is; it probably depends on the situation. But I think there are some dynamics you can end up in when you have a few friends who also notice the world is making that mistake, where the way to show how free-thinking you are, the way to show how rational, how independent, how agent-y you are, is to really, really not make the mistake that the world makes, really push in the other direction.

And I think the altruistic angle is one aspect of this, and this is maybe the most adjacent to some of the other things we’ve talked about in terms of moral perfectionism. I think there can be a version of this that might be a bit harder to hear if it sounds like it’s something that you do, and is closer to: you’re actually just playing the signalling game still, but you’re playing it with a different circle.

Luisa Rodriguez: You’re playing with people who are very rational, and you want to show that you are not blindly following conventional wisdom; you’re thinking it through for yourself. And you think that actually, most people have it wrong, but you’re smarter than them.

Alex Lawsen: Yeah. I think this can often come up in a context of biting philosophical bullets, where you’ve realised what I consider to be an important fact about the world, which is that most people have moral intuitions that end up conflicting in some way. So they will end up needing to notice that there is at least one unpleasant conclusion that they will have to reach. I think that’s an important thing to realise, and it’s a thing that most people don’t realise.

So maybe you’ve realised that, and you go, “Cool, I should probably bite one of these bullets.” And then there is just, “I bite all the bullets!” — enthusiastically biting them, or even maybe arguing really strongly that they’re not even bad, they’re like tasty bullets. What’s going on here? I have some hypothesis that what’s going on is the way to show that you are part of this group of cool people who has broken away from the sheep of mainstream society is to push really hard in the direction where moving somewhat in that direction away from mainstream society seems like clearly the correct thing to do. And you’re like, “Well, this is good, so more of it’s better. So I’ll move as far away from that position as possible.” And I don’t know, most extreme positions seem kind of wrong.

Luisa Rodriguez: It’s such a theme from this conversation: We notice something important, we try to correct for a mistake that we make, and then we correct too far and end up making a very different mistake with other consequences that are bad. Just seems like another clear case of it.

Are there other examples you want to highlight of ignoring conventional wisdom in careers?

Alex Lawsen: I don’t know, there probably are. Maybe you should just take away some message, which is: If you’re career planning, you should at least ask the question, “If you weren’t worrying about impact at all, what decisions would you be making?” I think you should know what that path looks like.

Let’s say someone told you that whatever you chose to do just turned out to be really great for others: you happen to get some luck potion, and whatever you choose turns out to just be great for the world. What do you really hope you decide to do? Or find some way of ruling out the “what’s best for other people” thing and just ask yourself what that path looks like, what strategies you would be considering if you were on that path — and then don’t instantly rule them out once you start considering the importance of helping other people as well. Probably your decisions will change. Probably it won’t look exactly the same. But “it won’t look exactly the same” and “it will look like the exact opposite” feel like pretty different claims.

Luisa Rodriguez: Yeah, that sounds totally right.

Doing community work even if you’re not suited to it [02:06:46]

Luisa Rodriguez: I guess a lot of people come to 80k advisors with the idea that they should do community building of some kind — for the field of AI, for example — and the thinking is that they don’t think they’re a good fit to work on AI directly, but they can still imagine having a big impact by convincing other people to work on AI. When is this line of reasoning a mistake?

Alex Lawsen: I guess maybe the thing that should be kind of obvious is if you would in fact hate community building, that’s a pretty good reason to think it might be a mistake. And I think this is a common feature of lots of the stuff I end up saying, where people are worried that maybe they should be doing something that they really don’t want to do. And I’m like, probably you shouldn’t be doing something you really don’t want to do.

Maybe the answer to “How do I know if it’s a mistake?” depends on a bunch of facts about you and a bunch of facts about the situation you’re in, and that’s why we do individual service. But I think some reasons that could cause people to make a mistake in this direction is this thing about “I really want to be doing the best thing right now” — rather than “I want to help other people as much as possible.”

The way it might come apart is: Clearly if you hate it, there’s a reason not to do it. If you’re struggling for time, and you don’t see your long-run path looking very much like it contains the same kind of skills as community building helps you develop at the moment, then I suspect that you’d be better served setting yourself up well for whatever that other long-term path is — even at the cost of a little bit of impact right now.

So maybe that’s a useful way of distinguishing when it seems like it might be a good idea: What kind of skills are you building while you’re doing it? What are you getting out of it? And if you’re not getting that much out of it that you expect to use later, then it might be a good idea — if it’d be really fun and if you have the spare time, then fine. But as soon as you’re like, “The reason I’m doing this is because I feel like I need to be having some impact, and I don’t see it being a long-term thing,” then my guess is that biting the bullet of like, “I’m not going to have quite as much impact this year. I’m going to set myself up well for having an impact later”: you’re allowed to do it on my lights.

Common themes [02:09:02]

Luisa Rodriguez: Great. So that’s everything we planned to talk about today, and I feel like I’ve gotten loads out of it. I’m around 80k, I hear lots of people talk about some of these things, and I still just learned a lot, and hopefully will make fewer of these mistakes myself in my career. So thank you.

But before we wrap up, I am curious if you had any reflections having done this, or any other mistakes that come to mind that we can cover before we call it?

Alex Lawsen: Thank you both for being up for chatting about it and sharing a bunch of your own experiences looking at some of these things. Yeah, I think I maybe noticed, as we were talking, a couple of themes that we kept coming back to. I’m not sure I’ll catch all of them, but it seems worth trying to do that.

Maybe one theme that seems worth pulling out is something like going too far in a good direction. I think lots of the stuff we’ve talked about has been like, “Conventional wisdom says this. You’ve noticed that conventional wisdom seems to not be tracking quite the thing you care about, or maybe even you just think conventional wisdom is wrong. And so you go, ‘I want to move conventional wisdom in this direction'” — and then you just move it way too hard in that direction. I don’t want to even make the general claim that you should never move things that far. I in fact think you should push some things pretty hard, and keep other things at a satisficing level. But I think there’s some amount of the correct update to make is not always like a really, really big one.

Luisa Rodriguez: To the other extreme. Yeah. That just makes sense, and it also makes sense that it’s hard, and so lots of people make these “mistakes.” If the optimal amount to move in a direction is a bit, but not all the way to the other extreme, finding the exact spot that’s ideal is just going to be really hard. And especially when you’ve got decisions as hard as these. We’ve already kind of qualified that people should not feel bad or guilty or ashamed if they’ve done some of these things. It’s just difficult because you’re doing this balancing act.

Alex Lawsen: I think it’s actually a dynamic of how language is structured, or how people speak to each other, that makes this hard to talk about. And I don’t mean hard emotionally; I mean we don’t quite have the right vocabulary for it. The thing that I want to point out here is we don’t have a neat single word for “I agree with you in the direction, but not as hard as you.” If people are estimating probabilities, then there’s still not one word, but there’s an idea. Let’s say you think there’s a 90% chance something’s going to happen, and I think there’s a 75% chance it’s going to happen. In this case, I can say I think it’s less likely, and it’s still kind of hard to explain why. If I explain why it’s less likely, but I don’t say any numbers, my guess is an observer that didn’t hear the start of the conversation isn’t really going to be able to tell whether I’m saying 75% or 10%.

Luisa Rodriguez: Right. It’s like, are you saying it’s less likely than I think and therefore you think it’s unlikely? Or are you saying it’s less likely than I think, it’s still probably going to happen, but you believe it less strongly than I do?

Alex Lawsen: Yeah, I think that’s exactly it. It’s just kind of clumsy and awkward to talk about situations where you and the person you’re talking to both basically think this thing on the left, rather than this thing on the right that everyone else thinks, but you’re a bit closer to the right than they are. There’s just kind of a difficult dynamic. And so I think part of what’s going on here is if lots of your friends and people you talk to about your career decisions are in a kind of similar place to you compared to conventional wisdom, it’s going to be hard for them to error correct. They don’t have quite the right vocabulary to error correct without just saying conventional wisdom at you.

Luisa Rodriguez: Yeah, that made me think that another thing that might contribute to this isn’t just the fact that you’ve realised, maybe correctly, that updating a certain direction is the right thing to do, but the more you’re in a community that has lots of people making that update. The more you might hear a bunch of people saying roughly the same thing — like, you should probably care less about conventional career wisdom than you might have thought before — and if like 20 people say it, each time you might get incrementally more extreme. When really it’s almost just like one very correlated data point that should make you update some, but less extremely than you might think.

Alex Lawsen: Yeah, and I think that is actually an important dynamic to talk about directly. Because, look, we’re both sitting here working for 80,000 Hours, and it just is the case that lots of people saying one-line things about careers have heard something like that one-line thing at some point somewhere on the 80,000 Hours website. So we do actually have specific reason to believe that there will be correlation in this place, if you and the four people you’re talking about this with all listened to the same podcast and got advice from me or someone else on my team.

I think that does seem worth pointing to as another part of this dynamic: not only is it hard to do the error correction, but maybe you hear the same thing five times, and you treat this as five independent pieces of evidence. Probably not consciously; probably if someone asked you how correlated did you expect these opinions to be, you’d say “very correlated,” but no one asked you that. So you just heard the five things and went, “Wow, really? Sounds like I’m underrating.”

Luisa Rodriguez: Right. “A bunch of different people think that this thing is underrated. I should update really hard.”

Alex Lawsen: Yeah. And then are there other themes that seem important? I’m really uncertain about how widespread this is, though I’m also uncertain about how widespread exactly these correlated-ideas dynamics actually are.

But I do think some of the things we talked about are related to moral perfectionism or feeling like a very strong obligation to help others. And I think the reason I’m pausing here is I think we both, as individuals, feel this a tonne.

Luisa Rodriguez: Definitely.

Alex Lawsen: And so I typical mind a bit when I’m talking to people: I sometimes think, “Wow, this person’s experiencing this thing. They must be thinking about it in exactly the same way as me.” And this probably isn’t true. So my best guess is that fewer than 100% of people who are making some of these mistakes, that we’ve labelled internally for us as a moral-perfectionism-adjacent thing, it’s probably not all of the reason, it’s probably not for all people. But my guess is it’s for more people than just us. And I think the episode that our colleague Rob did with Tim LeBon is like a good starting point if you’re interested in exploring this more. I got a lot out of it.

Luisa Rodriguez: Absolutely. I did too. Thank you so much, Alex. This has been great. I think people, well, I hope you get a lot out of it. I guess if you’ve made it this far, hopefully you’ve enjoyed it.

Alex Lawsen: Yeah, thanks. Thanks for having me on. I hope I said some sensible stuff. I’m kind of sitting here going, “Oh, wow, I said a lot of stuff. Maybe some of it’s true.” If we didn’t earlier say the mistake of, “You hear someone that sounds pretty confident and pretty smart on a podcast saying these are the conclusions you should reach, so you just go, ‘Cool, I should reach those conclusions.'” Don’t do that without evaluating them. If you think I’m wrong, I’m probably wrong.

Luisa Rodriguez: Relatedly, the advice for one person, you’ll want the opposite for another. So if you think Alex is wrong, if you think I’m wrong, it might also be because you happen to be one of the people who need to hear the opposite thing. And that is also fine and normal. But we’ll leave it there. Thanks again.

Luisa’s outro [02:16:55]

Luisa Rodriguez: All right. The audio engineering team is led by Ben Cordell, with mastering and technical editing for this episode by Milo McGuire and Dominic Armstrong.

Editing for this episode by a combination of myself, Katy Moore, and Keiran Harris — who also produces the show.

Full transcripts and an extensive collection of links to learn more are available on our site and put together by Katy Moore.

Thanks for listening.

Learn more

Know what you’re optimising for

Career guide landing page

Anonymous contributors answer: What mistakes do people most often make when deciding what work to do?

Anonymous answers: How risk-averse should talented young people be about their careers?

Related episodes

About the show

80k After Hours is a podcast by the team that brings you The 80,000 Hours Podcast. Like that show, it mostly still explores the best ways to do good — and some episodes are even more laser-focused on careers than most original episodes. But we also widen our scope, including things like how to solve pressing problems while also living a happy and fulfilling life, as well as releases that are just fun, entertaining, or experimental. Get in touch with feedback or suggestions by emailing [email protected].

Subscribe here, or anywhere you get podcasts: