Hide table of contents

Gesturing at a thing to mostly avoid. My personal opinion.

This topic has been discussed on the EA Forum before, e.g. Free-spending EA might be a big problem (2022) and The biggest risk of free-spending EA is grift (2022). I also wrote What's Your Hourly Rate? in 2013, and Value of time as an employee in 2019. This piece mostly stands on its own.

There's a temptation, when solving the world's toughest and most-important problems, to throw money around.

Lattes on tap! Full-time massage team! Business class flights! Retreat in the Bahamas! When you do the cost/benefit analysis it comes out positive: "An extra four hours of sleep on the plane is worth four thousand dollars, because of how much we're getting paid and how tight the time is."

The problem, which we always underindex on, is that our culture doesn't stand up to this kind of assault on normalcy. No altruistic, mission-oriented culture can. "I have never witnessed so much money in my life." [1]

What is culture? I often phrase it as "lessons from the early days of an org." How we survive; how we make it work despite the tough times; our story of how we started with something small and ended up with something great. That knowledge fundamentally pervades everything we do. It needs upkeep and constant reinforcement. "It is always Day One" [2] refers to how Amazon is trying hard, even as they have grown huge, to preserve their culture of scrappiness and caring.

What perks say

Fancy, unusual, expensive perks are costly signals. They're saying or implying the following:

  • Your time is worth a lot of money
  • You are special and important, you deserve this
  • We are rich and successful; we are elite
  • We are generous and you are lucky to be in our orbit
  • You're in the inner ring; you're better than people who aren't part of this
  • We desperately want to keep you around
  • You are free from menial tasks
  • You would never pay for this on your own—but through us, you can have it anyway
  • We're just like Google!

Some of these things might be locally true, but when I zoom out, I get a villainous vibe: this story flatters, it manipulates, it promotes hubris, it tells lies you want to believe. It's a Faustian trade: in exchange for these perks you "just" have to distort your reality, and we're not even asking you to believe hard scary things, just nice ego-boosting things about how special, irreplaceable, on-the-right-track we all are.

Signals you might want to send instead

The work cultures I prefer would signal something like the following:

  • We're normal people who have chosen to take on especially important work
  • We have an angle / insight that most people haven't realized/acted on yet
  • We might be wrong, and are constantly seeking evidence that would change our minds
  • We should try to be especially virtuous whenever we find ourselves setting a moral example for others
    • (We aren't morally better by default, although we may have a bit more information; we are always learning as we go)
  • We are focused on the long term good for the world
  • Tons of people are suffering / will suffer in the future
  • We remind ourselves of this regularly
    • (We regularly acknowledge suffering in the world; we definitely don't do things that are cushy)
  • We invest a lot in our people and their relationships, and act to preserve that value
  • Everyone works hard, sometimes they have to do things they don't want to do
  • We are in it for the long haul so don't burn out
    • (We might use money in various ways—to make our work relationships stronger or stave off burnout—but we aren't profligate.)

Instantiations of this culture will vary a decent amount—but I expect that a lot of altruistic orgs have/want to promote at least some subset of these values. Fancy perks push (perniciously, temptingly) against them, eroding our culture.

What to do

I don't have a prescription for what to do instead—the best ideas are certainly going to be specific to your team or organization's goals and existing culture.

The first and most fundamental question is, should you pay people well, and if so how well? Really tough question. I don't have a strong general opinion about this. The only clearheaded point I can make is that overpaying and underpaying both send powerful signals; sometimes those signals are what you want to be sending and other times they're really not. Carefully consider what you're signaling with your pay structure, e.g., high salaries might say: "everyone is equal" or "we hire only the best" or "we expect you are compromising your values to work here". And low salaries might say: "everyone is equal" or "we're all here because we care so much" or "we don't value your work".

I'd suggest making a cultural distinction between "perks" and "business expenses". I can't think of too many cultural reasons to limit business expenses; in fact it causes other problems if there's too much paperwork around getting reimbursed for basic things like equipment, transportation and services. (What counts as a perk? Business class travel in most cases is a perk, but a restaurant might be a business expense or a perk depending on the context.)

One trick to try is pushing the decision making about a perk out to the recipient. Maybe bringing a massage therapist into the office is a good idea, but try charging the employee for the appointment instead of making it free. That blunts the "perkiness" a lot, in a way which might annoy people in the short term but doesn't have most of the drawbacks of free perks. (I have seen this work particularly well in combination with slightly higher-than-usual salaries, e.g., to make a cultural point that we pay you well so you can make your own choices about perks.)

If the concept of carefully designing a culture is new to you, read this short post: Don't Fuck Up the Culture by Brian Chesky. To design my own company's culture, I talk (continuously) with my cofounder and the rest of the team about what works well about what we do, what surprises them; then I write those things down, and try to take it to logical extremes and think about the edge cases. Expect and embrace that your cultural choices have trade-offs. Leaders should exemplify the culture in their day-to-day.

For frugality in particular, I think you could land anywhere from "ultra frugal, no unnecessary stuff" to "spend money in the best interest of the org." If you might be perceived as a leader or a visible example, pick a point on the scale that you can see yourself personally exemplifying for a long time.

A couple examples from my company Wave

Not too long ago, Wave would pay for an upgrade to business class travel if the upgrade was less than the cost of the economy ticket (which was often possible for flights to Africa using the bidding-based systems of many airlines). We stopped paying for these upgrades and this seems like a pretty clear win to me.

We also ran all-expenses-paid "retreats" in Africa, in very nice hotels, for our remote team. These were costing upwards of $3500 per person for a 5-day trip. The biggest problem with retreats was the cost and perkiness; but the second biggest problem in my view was that attendees weren't getting a clear picture of the lives of the people we were ostensibly trying to help. We've basically stopped having retreats at all, which to me is not the greatest outcome, so I am still trying to figure out what to do instead that feels more frugal while getting our team a better sense of what life is like for our users.

We currently pay $550 per month for a coworking space membership per person for our remote team, which seems high and perks-y to me; some people probably benefit from coworking quite a lot though and their work could be greatly impacted by reducing this too much so it's a tougher decision that I'm working through right now.

For our remote team, we cover whatever home office equipment they need, with spending guidelines-but-not-limits, with the hopes that people who need expensive equipment can feel the license to get it and those without particular needs will be nudged towards cheaper options. (I don't really know how well this policy is working today though, and we are considering changing/simplifying it.)


(From early readers, or inspired by the comments on Free-spending EA might be a big problem.)

Q: Spending lots of money can be a positive signal to some: it says you're willing to take reasoning seriously and put your money where your mouth is. If that is culturally important, can we keep that benefit?

I think we can preserve the benefit by choosing targeted ways of using money wisely. If you think that it's very important to pay people highly, for example, then do that loudly and transparently—but get rid of the other perks. Basically, choose your spending wisely, explain your reasoning for it, and make it clear that it's an exception to an otherwise-limited budget.

Q: Excessive frugality can lead to distraction, losing sight of the bigger goal in order to save a bit of money. For example, if you encourage people to cook their own food ahead of an event instead of feeding them there, aren't they going to waste valuable time doing that work?

I agree with this concern and if you made me choose I would usually choose to offer free food! However, have you considered charging a nominal fee for the food? All things considered, any choice here seems reasonable (free food is hardly shocking), so maybe think about it and pick the culturally best one, and if you do give free food, make the quality of the food line up with the rest of the message you're trying to send.

Q: Does it make sense to pay more for a fancier hotel if it has substantially better Wi-Fi and the person might do some work in the room?

I'm worried about this one. In most parts of the world, there's a better way to get cheap good internet than paying premium hotel rates for it. I would like to see us goal factoring this type of decision more. Fancy hotel rooms definitely feels like a pernicious perk to me. Maybe you've thought about it a lot and decided it was worth it, but this is one of the examples I'm trying to push back against with this post.

Q: I've gotten large benefits from hiring cleaners, ordering food, paying money to alleviate bureaucracy hurdles, etc. Do I have to give all that up?

No, those may be sensible expenses—but these are mostly individual lifestyle choices. One of the things that I notice about "perks" is that they tend to shift individual lifestyle choices about how to spend your personal money into org-level decisions. I'm advocating for lots of these decisions to stay at the individual level rather than standardizing on them.

Q: Perks are tax efficient: paying someone $150k plus $50k in perks is more 'take-home value' than straight-up paying them $200k. How should I think about this?

It's a small factor. I don't think you should let tax efficiency override a potentially high-value cultural decision. I would usually rather pay more than offer more perks but I can see the tradeoff going other ways in some situations.

Q: How has your (Lincoln's) opinion changed since your 2019 blog post, Value of time as an employee? This seems to be pushing in the opposite direction.

Interestingly, I still find myself mostly agreeing with what I wrote in that post. In 2019 I saw instances in my own company of people not feeling free enough to spend money on things that were important. In 2023 I see instances (both at Wave, and in EA) of over-spending. My 2019 post has a bunch of references to the cultural impact of visible spending, which is the whole point I'm trying to make here too, so I'm glad to see it there :). One thing from that piece I disagree with now is: "the company should basically always be willing to pay for any time savings that is cheaper than the person’s hourly wage". I no longer think this is true, for cultural impact reasons, and I recommend being careful about what you visibly spend on. The 2019 example of spending money on coffee shops for employees is also something I would be inclined to argue against today, because of the third option: there might be ways to get the employees to pay for coffee shop visits on their own if they benefit from them, rather than subsidizing them.

Thanks to SM, RH and TF for comments on drafts.

  1. This post is not meant as an attack on FTX specifically, although they are a highly relevant example, but I see it in lots of places including Wave. ↩︎

  2. I don't endorse everything about Amazon's culture, but the Day One idea is particularly relevant to the cultural point I'm trying to make. ↩︎





More posts like this

Sorted by Click to highlight new comments since:

I think this principle is important, but question using Amazon as a positive counter example, and I think the reasons why are indicative of why this is hard.

My Amazon info is out of date. But back in 2015 I had many friends there, and it was a miserable place to work. The at-desk crying was past its peak, but stupid applications of frugality were so common they had a word for it (frupid).

Highly paid developers were given underpowered laptops that slowed down their work. The savings on the laptops were clearly OOM less than the cost of the developers time (even though Amazon paid less than other bigtech), but they wanted to send a message to employees. They would occasionally spend more money for a worse solution because it signaled frugality, like door desks that cost than real desks.

And that's for highly paid developers. Amazon requires warehouse employees leave all valuables in cubbies rather than bring them on the floor, but won't buy locking lockers for them.

The overwhelming message from Amazon is "fuck you, we will hurt our own interests rather than make your life easier".

Maybe this level of malice is necessary to maintain mission alignment. Maybe it's necessary with programm... (read more)

I appreciate this post, thanks for sharing it! 

I disagree with — or feel pretty unsure about — some notable parts/interpretations of the post, and that's what this comment is focused on.[1]  But I agree with some other parts, especially the specific suggestions for organization-level policies, and really like this point: 

One of the things that I notice about "perks" is that they tend to shift individual lifestyle choices about how to spend your personal money into org-level decisions. I'm advocating for lots of these decisions to stay at the individual level rather than standardizing on them.

I also quite like the suggestion to make "a cultural distinction between "perks" and "business expenses" and the nudge towards transforming some ~perks for employees into services (asking people to pay, and then potentially paying more in salaries). (I should note that I haven't thought about this a lot.)


Disagreements with a broad interpretation of the message of the post (something like: "~be more frugal / avoid perks")

This feels like a "reverse any advice you hear" situation, potentially: one in which different people are drawn to different/opposite pitfalls, and advice targ... (read more)


I'm confused about what this post is suggesting should be the principle that determines which perks are acceptable and which are not. For example, you say "(We might use money in various ways—to make our work relationships stronger or stave off burnout—but we aren't profligate.)" What counts as profligate here? Everyone already agrees that we shouldn't be profligate, but I take it that this just means that we agree that the benefit of the perk outweighs the cost. Is it ever worth it to pay for a perk in order to save people time? You say that business class flights are never acceptable. What about getting flights that take 5 hours longer in order to save £100?

In the Oxford office, we have a small gym, on-site vegan caterers, snacks, drinks and a nap room. I would guess that you think these are unacceptable perks? Firstly, people in the Oxford office by my observations, work extremely hard, especially relative to the non-profit sector. I don't see much indication that this is leading to a cushy lifestyle and lack of seriousness. Secondly, the nearest gym to the office is about a ten minute cycle, so not having a gym would cost about 2.5 days per person of possible work time (assumin... (read more)

My guess is that this post is implicitly aimed at Bay Area EAs, and that roughly every perk at Trajan House/other Oxford locations is acceptable by these standards.

Perhaps worth clarifying this explicitly, if true—it would be unfortunate if the people who were already most scrupulous about perks were the ones who updated most from this post.

I think it would be good if the OP would clarify which office perks he is criticising. Perks vary a lot across offices - probably more generous at Google and hedge funds, less at Amazon, less at a paper merchant in Slough, not great for an academia etc. The terms 'normal', 'usual', 'nice' are doing a lot of work in this post but are never defined and I don't know what they mean. Some things are normal in offices (dishwashers, standing desks) but are also nice. 
Vaidehi Agarwalla
What's the difference between Bay Area spaces and Trajan ? They seemed roughly the same to me?
Maybe this is somewhat indicative of Bay spending culture?

I don't know if this is right, but I take Lincoln to be (a bit implicitly but I see it throughout the post) taking the default cultural norm as a pretty strong starting point, and aiming to vary from that when you have a good reason (I imagine because variations from what's normal is what sends the most salient messages), rather than think about what a perk is from first principles, which explains the dishwashing and toilet cleaning.

Reminds me of C.S. Lewis's view on modesty

The Christian rule of chastity must not be confused with the social rule of ‘modesty’ (in one sense of that word); i.e. propriety, or decency. The social rule of propriety lays down how much of the human body should be displayed and what subjects can be referred to, and in what words, according to the customs of a given social circle. Thus, while the rule of chastity is the same for all Christians at all times, the rule of propriety changes. A girl in the Pacific islands wearing hardly any clothes and a Victorian lady completely covered in clothes might both be equally ‘modest’, proper, or decent, according to the standards of their own societies: and both, for all we could tell by their dress, might be equally

... (read more)

The default cultural norm varies a lot across offices within countries. Should we anchor to Google, hedge funds, Amazon, academia, Wave, Trajan House, the nonprofit sector, the local city council  etc? So I don't understand which cultural norm the post is anchoring to, and so I don't understand the central claim of the post. 

One of the examples given in the post is the implicit judgement that EA doesn't want to be like Google - Google is an extremely successful company that people want to work for. I don't get why it is an example of excessive perk culture. It's true that FTX had excessive perks and also committed fraud. Google has nice perks but hasn't committed fraud. 

While there may be some perks in EA, it is also the case that work in EA is (a) extremely competitive and (b) highly precarious. Most people struggle to get jobs or get by on one year contracts, and have to compete for jobs with assorted Stakhanovite super-geniuses. This is very different to the rest of the comparably cushy nonprofit sector.  

While at times it appears the OP is arguing for the default cultural norm, he also says various things which seem/are inconsistent with that such as that we... (read more)

A small note The OP didn't spell our why menial tasks could be valuable, so here's why I think we should keep doing at least some menial tasks. I'm not saying we should eschew technology or other help (I'm not ideologically against dishwashers or hiring cleaners) but I think there can be deep value in spending a not-necessarily-very-large amount of time on some menial task - It can be a humbling experience, helping us identify with those less fortunate than us both in our own country and abroad. - Maintaining our own spaces (dishes, toilets etc.) can be a community and maturity building experience, through forcing people to spend extra time in community spaces, working together to solve problems unrelated to their work and to help each other understand each others preferences and rub up against each others expectations and cultural norms a bit. Obviously it can also cause conflict and stress as well which can be a tradeoff. - While we are doing practical things with our hands, we can rest and quiet our minds, which may sometimes lead to unexpected creativity or discovery. - They can help us appreciate more the other more cerebral work we might be doing - We can demonstrate to others looking in, and to our other workmates that we don't consider ourselves "elite" or "above" certain tasks

Q: Perks are tax efficient: paying someone $150k plus $50k in perks is more 'take-home value' than straight-up paying them $200k. How should I think about this?

It's a small factor. I don't think you should let tax efficiency override a potentially high-value cultural decision. I would usually rather pay more than offer more perks but I can see the tradeoff going other ways in some situations.

Being more explicit about how this interacts with another downside of perks, where some people will take the perk who don't actually benefit that much from it. A toy example:

  • You decide to offer free high-quality coffee ($5/serving)
  • Employee A loves this coffee and gets the full benefit
  • Employee B isn't picky, and would be very nearly as happy with cheap coffee ($1/serving)

You're spending $10 to effectively give $5 to A and $1 to B tax free. The overall tax rate is about 40% (federal + state + payroll; plausible), so this example is a wash. Real examples can go either way, depending on the costs involved and how many A-like vs B-like people you have.

I'd suggest that the relevant tax laws about perks make inefficiencies more likely here. While the topic is complex, optimizing for perks that don't generate taxable income imposes a constraint on what can be offered. For instance, business-class airfare and company retreats shouldn't generally trigger taxable income to the employee, but are likely to be worth significantly less than cash to the employee. It's been over fifteen years since I took tax, but it's not obvious to me that regularly offering $5/serving coffee would be a de minimis fringe benefit. And as a policy matter, it should be taxable -- it's too close to being disguised compensation.
Jeff Kaufman
For the particular example of coffee, that's my guess at what Google was paying per serving when I worked there (barista etc). Am I wrong in expecting that they would be being pretty careful about this?
They have actual tax lawyers, so no need to amend your returns!  That much of the cost was for a barista, etc. rather than the coffee itself is potentially relevant -- both from a tax standpoint and for assessing your argument.  One could pitch the devolution of coffee-producing labor to someone making a tiny fraction of what the average Google employee makes as primarily for the benefit of the employer -- it gets the employee back to their desk more quickly. That makes it less likely to be seen as "disguised compensation." Presumably your hypothetical $1 coffee solution was DIY? That being said, I do not drink the stuff so could be making some assumptions that are off-base.
Jeff Kaufman
In practice people would wait in line for the barista, sometimes for quite a while, and the baristas weren't much faster than doing it yourself, so the time argument seems to point the other way?
Sounds like they should have hired more baristas. Did a little updating: the tax treatment of employee meals on-site is a bit different than other perks, with some changes since I took tax. They are in some circumstances not tax-deductible as a business expense by the employer or taxable to the employee, so the IRS gets its fill a different way where the employer is not exempt from income taxation. E.g., this discussion of a 2019 IRS technical advice memorandum or TAM (also this). So going back to your hypo, if that is the tax treatment in a particular situation, there may not be a tax advantage depending on whether the employer is a 501(c)(3). Traditionally, coffee and doughnuts are excludable as de minimis fringe benefits, but as the first linked article explains the reason for that is that the value of each snack event is small and accounting for the benefit is impractical. I'm not sure that would hold for a barista-run coffee shop where the employer could scan the employee's badge and track usage that way. A study of the 50-page TAM opining on one employer's on-campus dining might help but doesn't seem very high-value.

Thanks for writing this out and distinguishing between workplace perks and personal spending.

This movement is always going to have a cultural clash between the social pressures of class status and the demandingness of charity as explained by philosophers like Parfit, Singer, and Unger. Lifestyle creep is pernicious, and EAs are excellent at rationalizing luxury spending. SBF's billionaire lifestyle is an extreme example.

It seems that many EA orgs are modeling themselves after bay area tech companies which are peculiar elite workplaces with massive compensation packages that tend to wrap around and cater to the employee's entire life. That is not the norm in other high productivity analogous workplaces, like in academia, labs, or non-profits.

I like the proposal of having employees pay for any perks they enjoy beyond what is strictly necessary for job function (computer, workspace, healthcare/retirement in the U.S.). Maybe occasional lunches and a coffee machine. If people think an expensive perk makes them more productive - great! Buy it yourself, and if it really does make you more productive, you'll get rewarded for it at your performance review. It's not hard to have people pay for on site perks.

Maybe occasional lunches and a coffee machine


Flagging that I think lunches every single day may be a great productivity investment for an org to make – the alternative is likely people leaving the office to find lunch, taking a significantly longer lunch break. The cost is only several dollars a day per employee, versus a benefit of perhaps several times that. Ditto for having good coffee and snacks available at all times.

Vaidehi Agarwalla
For reference - the cost of meals in the Bay is $15 - 30 per meal per person. So something like $450-900/month per person (not including ops overhead)
For an individual meal or if buying ingredients in bulk and doing buffet style?
Vaidehi Agarwalla
For catering.
Do people take lunch breaks solely or predominately because they need to eat, or is the need to take a break (and/or get out of the office for a bit) an important motivation as well? To the extent an employee's lunch break is serving the latter purpose, providing quickly accessible in-house food could be partially displacing that need to a different occasion.
If people want to get outside and clear their head, presumably a short walk is generally better than waiting in line at a lunch spot? Regardless, my sense is making food available for those who want easily available food is good idea.
The alternatives also include people packing their own lunch, or having people pay for the lunches they buy on site (a small cafe for example). If the worksite is in a downtown area, there are excellent options within a few minutes walk, and people can pick exactly what food they want. If people are worried about the loss of a few minutes walking, or they are are far away from food, they can order food delivery. Free work lunches are not a necessary perk, and people can easily pay for it if they think it's worthwhile. The exception I'd make is someplace like a K-12 school where they are already giving kids (free) lunch, and marginal costs are minimal to extend the benefit to adults. Or if the work is in the food industry.
Packing your own lunch takes time (similar to going out to lunch), and charging everyone for lunch each day would make the culture between employer/employee feel more transactional (which seems bad for morale).

I wonder how much a person's view on charging for meals as morale-reducing is affected by exposure to perky vs non-perky cultures. For most US employees, the idea of your employer buying you lunch everyday is not even on their radar.

I don't understand why a response like this would get so much bad Karma - its a reasonable response in good faith - I think we perhaps need to get better at separating disagree and bad karma voting.
Love this concise argument and reference to philosophy and other types of orgs. Nice one!

Most of my experience is in the AI Safety sphere, and for that, I think perks and high salaries are critical. I'd love to see Alignment orgs with more of these things. The issue is we need high talent. And that high talent knows their worth, especially right now. If they can get Business Class working at Meta AI, I'd want to offer them First Class. If you have the money to make it happen, outbidding talent and being a talent attractor is important. Perks signal a job is high status. Retreats in luxurious locations signal high status. High status attracts high talent. I can't ask everyone with great talent to work on safety just out of the goodness of their hearts.

I'm interested to know if you think there could be a problem with attracting people where remuneration is the biggest decision-making factor.

I am somewhat sceptical of this argument (I have seen it used to say "we should pay minimum wage because if you really care you'll take the job anyway") - but I also wonder if we can sustain a bidding war against e.g. DeepMind for talent, and if a better approach might be something like GiveWell's ('salary shouldn't be the primary reason people leave GiveWell, but also not the primary reason that they join').

What do you think?

I don't think anyone can win a bidding war against OpenAI right now, because they've established themselves as the current "top dog". Even if some other company can pay them more, they'd probably still choose to work at OpenAI instead, just because they're OpenAI. But not everyone can work at OpenAI, so that still gives us a lot of opportunity. I don't think this would be much of a problem, as long as the metrics for success are set. As mentioned above, x gains in interpretability is something that can be demonstrated, and at that point it doesn't matter who does it, or why they do it. Other fields of alignment are harder to set metrics for, but there are still a good number of unsolved sub-problems that are demonstrable if solved. Set the metrics for success, and then you don't have to worry about value drift.
I wonder (with no data or inside knowledge to back it up) if this mindset might be part of the reason why labs like OpenAI and perhaps to a lesser extent anthropic have swung towards capabilities and away from safety. People who are primarily money focused might be more likely to push an organisation towards capabilities because that's where the big money is at - not AI safety.  I instinctively don't like the idea of "High status" jobs as well. I think EA can do better than having to signal jobs as "high status" to achieve what we want to achieve. Community matters and reputation matters, I don't think its healthy to have a community where orgs feel like they need to sell out to greed to get their work done well and I'm not sure funders will necessarily feel great about that either in the long run. There's "value drift" and then there is "value compromise" and it seems like there's the risk of diluting the principles of EA in a race for the bottom to get the best AI researchers working on safety.  As a side note I'm not sure we'll ever have the finances be able to sustain competing for a long period of time on salaries and perk with orgs like Meta AI even if we tried. "I can't ask everyone with great talent to work on safety just out of the goodness of their hearts."  You might well be right here, although like Jack says below people can work for a decent salary while still be doing work for the right reasons. I think we don't want everyone with great talent but I do understand the argument
So I think there's a huge difference between other EA causes and AIS. You can probably accomplish a good number of other objectives in EA without these, but I still think trying to make them higher status might still be useful. It's a way of signaling what's important in society and what's valued. If I knew a way to make someone working in Pandemic Preparedness as high status as the NBA, I probably would. That being said, AI is a different beast. Places like San Francisco are filled with people working 70+ hours a week, hungry to get ahead in someway in AI. I'd love to tap into that hunger, with the metric for success being Alignment. It would need to have actual metrics for success, though, like provably solving certain certain aspects of the problem, or making a huge discovery in interpretability. If someone can accomplish demonstrable huge gains in this, I don't really care what their personal motivations are.
Thanks for this added explanation, makes more sense when you put it that way. Again I can never understand somewhere like San Franciso, living in sleepy Northern Uganda and working 40ish hours a week with probably average productivity by your standards! I also don't necessarily care what their motivations are, I'm just dubious whether the motivation of greed is going to achieve those huge gains you are hoping for. My 2 big questions hre are 1. Can we actually raise enough money to make the EV of the greedy talent worth it, vs. others who are still very talented but willing to work on safety for less? Like if we hired 2 EA safety people at 100k, would the greedy one at 200k really be achieving more? 2. Will either the org or the individual not eventually be corrupted towards capabilities by the big money anyway? Either when our EA money runs dry or they just get offerred more than we could (300k rather than our 200 whatever it is). Basically what Jason said below. "Also, if a lab has a massive payroll to meet, that fact alone might subtly push its leaders toward what will generate revenue and/or attract profit-motivated investors., as those revenue sources are more stable/reliable than charitable grants mostly from a single donor." Its a reasonable argument you make though, I'm fairly uncertain about this.
1. I would have said no a year ago, but a lot of people are now much more interested in AIS. I think there's a lot of potential for much more funding coming in. The binary greedy vs. non-greedy human sounds strange to me. What I can say is many EA types have the mentality of neglectedness, how they can individually have the most impact, etc. Many EAs would probably say they wouldn't be working on the things they were working on if enough other people were. This is great in isolation, and a mentality I usually hold, but it does have problems. The "greedy" humans have the mentality of "someone else is going to do this, I want to get there first." Individually, this doesn't change much. But if you multiple people doing this, you get people competing with each other, and usually they push each other to get to the outcome faster. 2. Yes. But everyone's pushing hard on capabilities right now anyway. This has always been a problem in AIS. But we can't really do anything without running into this risk. But I think there's a big difference between employees at an org, and people starting orgs. I'd be fine with existing orgs attracting talent the way I mentioned, but I wouldn't want to throw money at someone (who's only interested in status) to start their own org. It's certainly tricky. Like, I can imagine how the leaders of an org can slowly get usurped. Holding current leaders in AIS in prestige can possibly mitigate the risk, where people with senior status in the field can function as "gatekeepers". Like, a young physicist who wants to gain clout, only for the sake of their own status, is still going to have to deal with senior members in the field who might call bs. If enough senior members call bs, that person loses status.
Also, if a lab has a massive payroll to meet, that fact alone might subtly push its leaders toward what will generate revenue and/or attract profit-motivated investors., as those revenue sources are more stable/reliable than charitable grants mostly from a single donor.

I agreed with most of the beginning of the post, but the specifics of where to cut perks seemed highly context-specific and I think readers should beware that.

For example, in the Bay, a lot of community “perks” like the former Lightcone coworking space are much less perky than they may appear because people live in rented rooms in group houses and don’t have stable jobs. They made the community in the Bay much more possible. High salaries are often a must not only because of high cost of living but to compensate for not offering benefits and low job security.

I guess what I’m saying is let’s not stigmatize the appearance of nice things when there are lots of tradeoffs in different org/community circumstances.

This resonates. As a minor but pretty unambiguous example, I found it uncomfortable that at EAGL there was basically unlimited free wine. That seems like it would have cost enough to save at least a couple of AMF-lives, or allowed a few extra people to attend the conference, or whatever counterfactual seems apt, and it's hard to imagine what version of the 'it marginally improved my productivity' argument would be convincing.

Thanks for this fantastic post I love it!

Even though this has been discussed a reasonable amount., I think there's continuous need for discussion and iteration around topics like perks, salaries and work cultre.

I think on the work retreat front there are so many options which are far cheaper than the bougie hotel one for remote workers that you could explore. Perhaps hire a simple but nice Air B&B for the team near the beach, or find a simple girl guide/boy scout retreat center which can host people really well, and at a fraction of the cost often with nice outdoor and indoor activities around (table tennis/pool/ropes course/river etc.) 

My point of disagreement will be on the salaries which I think should also be lower, but that's kind of aside from the article's main point. I also think (although not sure) I find myself in the minority on that front on the forum here as I seem to take both a karma (which I don't understand) and an agreement (which I do understand) hammering when I mention it ;).. However I LOVE your comment here about salary tradeoffs

"I don't have a strong general opinion about this. The only clearheaded point I can make is that overpaying and underpaying... (read more)

I'm gonna push the general line on this and say "it's also the job of individual orgs to negotiate their salaries with individual people". I have time for a discussion of vibes but I think markets outcompete central planning. 

Likewise, there is clearly another bad equilibrium where orgs spend roughly the same amount of money on worse outcomes because they are scared of looking bougie. 

Hard problem.

This market logic doesn't apply when essentially all of EA's money comes from one source. These aren't companies with diverse revenue streams from product sales. Open Phil can tomorrow make every EA org pay less by giving each org less money, if they want.

Or probably even by writing grant terms in a certain way...
That too. I find the widespread reluctance to point the finger at OpenPhil for perceived problems in EA completely bizarre. Whatever problem you think you've identified within or across orgs, I can almost guarantee OpenPhil can fix it because one way or another, they're funding it. That they haven't done so to date presumably indicates they don't think it's a problem or they simply don't care.

Yeah a significant consideration for me in whether to be less professionally involved in EA is exhaustion from centralized funding and the weird power dynamics that ensue. I would rather build products that lots of people can use and lots of investors or donors would find attractive to give money to than be beHolden to a small coterie of grantmakers no matter how well-intentioned.

I don't think that's quite fair. There are good reasons for major funders to be hesitant to use the funding hammer to micromanage grantees. One could conclude in some cases that there is a problem but that trying to fix it with the funding hammer poses too many downsides.

I don't think anyone is advocating for centralized planning (which would be generally illegal anyway). But I do think an org-focused compensation strategy poses serious risks of drawing talent to the best-funded org rather than the org at which their marginal impact would be highest. In the for-profit world, the scrappy upstart can offer equity to compensate for lower base pay, but that's not a thing in the non-profit world.
Nice one Nathan I agree that markets outcompete central planning, hence why I think the purpose of the discussion is as much to sway the market a little through shifting the views of those EA types applying for the jobs - for example if I managed to convince a bunch of people that lower salaries were the best way forward then the micro EA job market would shift. I have asked for lower salaries before a couple of times, but I doubt that's a common practise.

Thanks for writing this! I'm not sure how I'd feel if orgs I worked for went more in this direction, but I did find myself nodding along to a bunch of parts (though not all) of what you wrote.

One thing I'm curious about is you have thoughts on avoiding a "nitpick" culture, where every perk or line item becomes a big discussion among leadership or an org, or the org broadly - that seems to me like a big downside of moving in this direction.

Just because, things I especially liked:


We should try to be especially virtuous whenever we find ourselves setting a moral example for others

(Though sometimes/often I think the excellent thing to model for others is "yes, I am really going to do this weird / not-altruistic-looking thing because it is the right thing to do)

2. Bringing in services to make them convenient but then asking people to pay sounds like a bit of a boondoggle but also really clever - I don't think I'd encountered this kind of compromise before. I'd be interested in more of this form!


Thanks for writing this. It resonated with some of my feelings.

One thing I've noticed about getting a lot of perks is, for some reason, it makes me feel "safe"; in a way that a high salary doesn't. Probably because I would just save most of the high salary, but perks force me to live like a rich person. And therefore, make me feel like a rich person. This, therefore, makes me feel that my job security/life is better/safer than it actually is. I would always prefer cash compensation. Cash doesn't distort my thinking (as much), and it helps me weather possible unemployment. Therefore, I would be much sadder to lose a hi... (read more)

I'd be really curious what sort of impact perks have on the cost of an employee at typical EA orgs. Is the difference on the order of 5% or 30% of an employee's salary? I mostly feel like I don't have a sense as to whether cultural / signaling considerations or cost considerations are primary.

I'm specifically confused about business class flights being a no go at your company.

I work for the British civil service. We're not allowed to expense tea or coffee, except if we have an external visitor, or to fly within Britain unless we get special permission - but we are allowed to fly business class if it's a longhaul flight.

I think it's the sleep, or possibly the privacy while working, that tips the scales.

Anyways, surprising difference!

In the US federal civil service, business class is only allowed under very specific criteria (like a very long flight with a business need to be on duty soon after landing). (edit: https://www.law.cornell.edu/cfr/text/41/301-10.103 ) The crux between your employer and private sector could be volume discounting. The US Government, for example, gets massive discounts on authorized economy fares -- but because business class is rarely authorized, gets no discount on business class. My guess is that if the UK government authorizes business class on certain flights categorically, it's getting a steep discount -- creating a better value for the employer's business objectives per pound than the private sector can muster. Thus, it'd be less perk, more business expense.
Shakeel Hashim
I wonder how much this is a US/UK thing because of the types of flights people are taking. My assumption is that in the US the vast majority of flights are domestic, and I’d agree that business class just isn’t worth it on those planes (aside from the length of the flight, the planes are also not that nice!). The equivalent would be UK-Europe flights, for which business class definitely doesn’t seem worth it. But most UK travel in my experience ends up being very long haul, normally transatlantic — and on those, business class is clearly much, much better than economy because you get a lie-flat bed. And on a night flight (especially coming back from the US), that can be the difference between sleeping or not, and that then gets you an extra day of work when you land.

Interesting read. 

I have no clue what typical salaries or perks are in the EA community, and how they compare with similar jobs in similar locations, or how people react to them. So I won't even try to express an opinion on the content itself. But it shows a very healthy self-awareness that this discussion is actually happening! 

Ultimately, for most people, I would imagine the single biggest "perk" of EA work is to be able to work on something they really care about. And the downside to this is that they probably work far harder than they are required to, because they see this as more than just a job. 


Some of these statements seem ambiguous... it's unclear how you should vote here if you think we should pay market rate.  
Nathan Young
The tool is still in development. I take it you didn't see the edit button? How could that functionality be clearer?
Nathan Young
Stake of the discussion so far: For the uncertain ones, I suggest people add their own comments that offer clarification.
There's often a pretty big difference between market for industry, government, and academia -- but the differences depend on field and level (e.g., industry is well ahead in most tech fields) so it's hard to write a question
Robi Rahman
Agreeing with Larks: If I agree with the first sentence and disagree with the second sentence, how am I supposed to vote?
Nathan Young
The tool is still in development. I take it you didn't see the edit button? How could that functionality be clearer?
Curated and popular this week
Relevant opportunities