Note: This is a repost of my Substack article, “The Cost of a Closed Loop” (linked above), where I lay out what a year in EA’s AI safety x animal welfare space taught me about how impact is actually allocated here, who it leaves out, and who ultimately bears the cost. I have chosen to post under my Substack pseudonym because using my real name would make it easy to identify the institutions I discuss in this piece, and my goal is not to “cancel” them but to push them to make changes in how they screen and fund people.
The Cost of a Closed Loop
For a movement obsessed with maximizing impact, EA sure does leave a lot of it on the table — I’d know, some of it is mine.
Last month, an AI safety org leader sent me a message that immediately pissed me off.
The message itself was cordial: a project they’re supporting needed a graphic designer (specifically, a “graphic design volunteer”), and they thought I would be a good fit, since their project addresses a problem I’ve worked on myself. But it frustrated me for two reasons: (1) I am not a graphic designer, and (2) they know I am not a graphic designer because they have seen, used, and referenced my work… while rejecting me from every program they run.
If it were just the first thing, I wouldn’t be writing this — hell, I wouldn’t even be on Substack — and I probably would have agreed to help. I’m an AI product designer, so involving me at the final hour to throw some colors and charts over an otherwise finished product is like asking an architect to choose the paint on a house someone else built. That being said, visual design falls well within my purview, and I regularly offer these services pro bono to orgs doing impactful animal advocacy work, which, by all accounts, this one is.
And I swear this isn’t me being butthurt over a few rejections, either (though of course they sting differently coming from an org whose mission I actually care about). But this one rejected me three times in the last nine months — once per trimester, if you will — and at some point in this gestation of rejection, they delivered something: my independent work, in their fundraising materials.
Congratulations, I guess, to everyone but me. It’s a ploy!
And I only know about it because a friend shared one of the org’s investor-facing artifacts with me. On it, there’s a section describing a product I conceptualized and shipped, with a “learn more” link to its showcase on my personal website. This is a product they played no part in because, ICYMI, they rejected me from all their programs.
To recap: my work is good enough to raise money with. I’m good enough to donate my labor to teams funded by that money. However, I’m not good enough to work on one of those teams.
Well! I’m sure there’s a great (s)explanation for this.
And I promise we’ll get to it, eventually. But first, some background:
A year and a half ago, I hadn’t heard of the term “AI safety,” let alone considered its relevance to animal welfare. I’d been working full-time on AI products for a couple of years at that point, and I was hungry for a new project — ideally, one that could help protect animals. So, I went looking — for an idea, a sign, validation… I didn’t really know. And in my research, I came across something awesome: a hackathon just for animal advocates working on AI. Turns out the niche I was trying to carve out for myself not only existed but had an entire community behind it — a community rich in ideas and the resources to bring them to life. I was buzzing!
I did the hackathon, which led to a fellowship, where I got an idea, turned it into a product, was invited to speak about the product on a few forums, got great feedback, then got more ideas, spent time researching those, realized I needed to upskill if I wanted to see them through, applied to courses and mentorships and fellowships, and then… suddenly, the momentum stopped and the rejections started pouring in.
And they weren’t all from this org, to be fair. I was also turned away from a rudimentary AI safety course series — twice! The first time, I applied to (and was promptly rejected from) their technical course. They encouraged me to take their other course — the one for people with no technical background — then come back and reapply to this one. Ouch. But, fine. I’ll bite.
Two days later, they rejected me from that one, too. Apparently, I didn’t even have the background to consider getting a technical background. Meanwhile, certain people with comparable backgrounds and skills who entered this space at the same time as me were now being invited to build and facilitate and mentor the programs that I couldn’t even get into as a mentee.
Naturally, I started to think that I’m a stupid piece of shit with sewage for ideas.
Except, I know I’m not, because the people rejecting me were also the ones using me and my work to expand their reach. My face, plastered on their conference brochures. My project, shared with their investors. Invitations to speak, to volunteer, to connect with the teams they actually funded.
Okay, so if it isn’t my work, then maybe I just suck at the process itself — maybe I’m a weak applicant or a bad interviewer, and maybe the people who started with me but surged ahead are simply more qualified for the job.
But if that’s the case, then why did one program silently reverse its decision right after I responded to the rejection email asking for feedback? (I never got the feedback, by the way. Just a follow-up email asking me when I can start.) And why did I get picked for opportunities to do impactful work only once I ventured outside the EA-AI-animal-welfare bubble? An acceptance into a responsible AI design program, an award that’s now funding my next independent AI x animals project — and none of it thanks to the movement that allegedly exists to resource exactly this type of work.
Because in here, there’s no shortage of opportunity, if you can name ten people you know and convince all of them that you’re already an expert. You don’t have to be one — in fact, you couldn’t possibly be one on a topic you heard about a year ago. And you shouldn’t have to be, either! The whole point of these programs is to help driven professionals learn an emerging field fast enough to be useful in it… or so you’d think. What we actually see are the same names and faces, first as fellows, then facilitators, then founders.
“Okay, but you’ve literally just described networking. What’s wrong with that?”
I mean, a lot of things, but I’ll save my lukewarm takes about that for my group chat since I don’t ever plan on making a Substack comeback. What I will say is this: you can run a movement on who-knows-who, or you can claim to allocate by measured impact. But you can’t do both — unless, of course, you genuinely believe that impact lives inside a very specific group of people and their friends. And if you don’t believe that, well, then it stands to reason that the way you’re screening for it right now is inherently flawed.
Obviously, none of this is unique to EA. Cronyism plagues almost every movement; EA is just the only one that insists its decisions are made on calculations of potential impact.
Oh, don’t wait for me to solve it. I don’t have the technical background for this kind of thing, remember?
Still, if I had to guess, I’d probably say something about how it results in a closed loop of mostly men — vouching for themselves, funding themselves, and then labeling the results as ‘impact.’
Feel free to check my work, though. And since I’m already so far behind my peers, I’ll just go ahead and submit my extra credit right now for consideration:
A few weeks ago, I applied for a product designer job at another AI safety org — not the one that already turned me away (they made it pretty clear that they’re only looking for volunteers right now). A product designer with an AI safety portfolio, applying for a product design role within an AI safety company. Listen, I’m not saying I was entitled to the job, but there aren’t exactly swarms of people who fit the bill here. I figured I’d at the very least get an interview.
Nope. Automated rejection. So, I did what any self-respecting cog-in-the-machine does and went to the team page on their website, hoping to figure out who got the job over me so I could stalk their LinkedIn profile and feel even shittier about myself. I couldn’t find them, but I did find something way worse (which, ironically, made me feel slightly better in the moment): excluding the CEO and COO, they list 31 technical employees, two of whom are women. TWO!
And the org with the graphic design request isn’t all that much better, just smaller. They currently list 12 fellows for a program I was turned away from, and only two are women. Two out of 12 beats two out of 31, but neither even comes close to EA’s own gender distribution (71% male, per this survey), let alone gender parity, and that is a low bar.
Oooh, sexism in tech. Groundbreaking, I know.
But then again, this isn’t just tech, is it? It’s AI safety, a field that exists to ensure evolving technologies align with human values, not our worst biases. Yeah, good luck keeping those biases out of the algorithms when you can’t even keep them out of your hiring practices.
Okay, so maybe I am a little butthurt over the rejections. You would be too if you watched, in real time, people — men — who started at the same place as you get handed opportunities on promise alone, while your prototype couldn’t even land you an interview.
There’s a name for this, I learned. The prove-it-again bias: men are judged on potential, while women are judged on what they’ve already delivered, and then asked to deliver it again. I shipped a product that funders liked and effectively got told to piss off but stay on hold in case they needed a volunteer, while men who haven’t even graduated college yet — no industry experience to their name — effortlessly walked through the doors.
And the worst part of it all is that even now — even after being dragged through the mud by this movement — my first instinct is still to say yes to an org that didn’t want me but wanted my free labor. Because what I’m after is not a paycheck — I already have one of those, and it’s solid enough that animal welfare orgs get to see 10% of it every month. What I’m after is the opportunity to help as many animals, current and future, as my skills will allow.
Which is probably exactly where this movement wants me. “Set your ego aside.” “The mission is bigger than you.” Saying no here feels like I’m choosing myself over animals, because the cost of the closed loop isn’t my dignity; it’s potentially billions of their lives.
So, I’m still putting in the work. Still upskilling, still learning, and still building — but I’ve started weaving in some impact calculations of my own to make sure my time goes where it’ll count. And per my calculations, the expected value of repeatedly throwing myself at a closed loop is… approximately zero. Which must mean, then, that the maximum value I can create for animals has to be outside of the loop. And so from here on out, that is exactly where my ideas and I will live.
But don’t count on my math. After all, I don’t have a technical background.
Is EA’s AI safety work leaving impact on the table?
Link post
Note: This is a repost of my Substack article, “The Cost of a Closed Loop” (linked above), where I lay out what a year in EA’s AI safety x animal welfare space taught me about how impact is actually allocated here, who it leaves out, and who ultimately bears the cost. I have chosen to post under my Substack pseudonym because using my real name would make it easy to identify the institutions I discuss in this piece, and my goal is not to “cancel” them but to push them to make changes in how they screen and fund people.
The Cost of a Closed Loop
For a movement obsessed with maximizing impact, EA sure does leave a lot of it on the table — I’d know, some of it is mine.
Last month, an AI safety org leader sent me a message that immediately pissed me off.
The message itself was cordial: a project they’re supporting needed a graphic designer (specifically, a “graphic design volunteer”), and they thought I would be a good fit, since their project addresses a problem I’ve worked on myself. But it frustrated me for two reasons: (1) I am not a graphic designer, and (2) they know I am not a graphic designer because they have seen, used, and referenced my work… while rejecting me from every program they run.
If it were just the first thing, I wouldn’t be writing this — hell, I wouldn’t even be on Substack — and I probably would have agreed to help. I’m an AI product designer, so involving me at the final hour to throw some colors and charts over an otherwise finished product is like asking an architect to choose the paint on a house someone else built. That being said, visual design falls well within my purview, and I regularly offer these services pro bono to orgs doing impactful animal advocacy work, which, by all accounts, this one is.
And I swear this isn’t me being butthurt over a few rejections, either (though of course they sting differently coming from an org whose mission I actually care about). But this one rejected me three times in the last nine months — once per trimester, if you will — and at some point in this gestation of rejection, they delivered something: my independent work, in their fundraising materials.
Congratulations, I guess, to everyone but me. It’s a ploy!
And I only know about it because a friend shared one of the org’s investor-facing artifacts with me. On it, there’s a section describing a product I conceptualized and shipped, with a “learn more” link to its showcase on my personal website. This is a product they played no part in because, ICYMI, they rejected me from all their programs.
To recap: my work is good enough to raise money with. I’m good enough to donate my labor to teams funded by that money. However, I’m not good enough to work on one of those teams.
Well! I’m sure there’s a great (s)explanation for this.
And I promise we’ll get to it, eventually. But first, some background:
A year and a half ago, I hadn’t heard of the term “AI safety,” let alone considered its relevance to animal welfare. I’d been working full-time on AI products for a couple of years at that point, and I was hungry for a new project — ideally, one that could help protect animals. So, I went looking — for an idea, a sign, validation… I didn’t really know. And in my research, I came across something awesome: a hackathon just for animal advocates working on AI. Turns out the niche I was trying to carve out for myself not only existed but had an entire community behind it — a community rich in ideas and the resources to bring them to life. I was buzzing!
I did the hackathon, which led to a fellowship, where I got an idea, turned it into a product, was invited to speak about the product on a few forums, got great feedback, then got more ideas, spent time researching those, realized I needed to upskill if I wanted to see them through, applied to courses and mentorships and fellowships, and then… suddenly, the momentum stopped and the rejections started pouring in.
And they weren’t all from this org, to be fair. I was also turned away from a rudimentary AI safety course series — twice! The first time, I applied to (and was promptly rejected from) their technical course. They encouraged me to take their other course — the one for people with no technical background — then come back and reapply to this one. Ouch. But, fine. I’ll bite.
Two days later, they rejected me from that one, too. Apparently, I didn’t even have the background to consider getting a technical background. Meanwhile, certain people with comparable backgrounds and skills who entered this space at the same time as me were now being invited to build and facilitate and mentor the programs that I couldn’t even get into as a mentee.
Naturally, I started to think that I’m a stupid piece of shit with sewage for ideas.
Except, I know I’m not, because the people rejecting me were also the ones using me and my work to expand their reach. My face, plastered on their conference brochures. My project, shared with their investors. Invitations to speak, to volunteer, to connect with the teams they actually funded.
Okay, so if it isn’t my work, then maybe I just suck at the process itself — maybe I’m a weak applicant or a bad interviewer, and maybe the people who started with me but surged ahead are simply more qualified for the job.
But if that’s the case, then why did one program silently reverse its decision right after I responded to the rejection email asking for feedback? (I never got the feedback, by the way. Just a follow-up email asking me when I can start.) And why did I get picked for opportunities to do impactful work only once I ventured outside the EA-AI-animal-welfare bubble? An acceptance into a responsible AI design program, an award that’s now funding my next independent AI x animals project — and none of it thanks to the movement that allegedly exists to resource exactly this type of work.
Because in here, there’s no shortage of opportunity, if you can name ten people you know and convince all of them that you’re already an expert. You don’t have to be one — in fact, you couldn’t possibly be one on a topic you heard about a year ago. And you shouldn’t have to be, either! The whole point of these programs is to help driven professionals learn an emerging field fast enough to be useful in it… or so you’d think. What we actually see are the same names and faces, first as fellows, then facilitators, then founders.
“Okay, but you’ve literally just described networking. What’s wrong with that?”
I mean, a lot of things, but I’ll save my lukewarm takes about that for my group chat since I don’t ever plan on making a Substack comeback. What I will say is this: you can run a movement on who-knows-who, or you can claim to allocate by measured impact. But you can’t do both — unless, of course, you genuinely believe that impact lives inside a very specific group of people and their friends. And if you don’t believe that, well, then it stands to reason that the way you’re screening for it right now is inherently flawed.
Obviously, none of this is unique to EA. Cronyism plagues almost every movement; EA is just the only one that insists its decisions are made on calculations of potential impact.
Hey, since you guys love math so much, here’s a problem for you: What is the impact of referral-based selection on a movement that’s already 70% male, given that three out of four people refer candidates of their own gender?
Oh, don’t wait for me to solve it. I don’t have the technical background for this kind of thing, remember?
Still, if I had to guess, I’d probably say something about how it results in a closed loop of mostly men — vouching for themselves, funding themselves, and then labeling the results as ‘impact.’
Feel free to check my work, though. And since I’m already so far behind my peers, I’ll just go ahead and submit my extra credit right now for consideration:
A few weeks ago, I applied for a product designer job at another AI safety org — not the one that already turned me away (they made it pretty clear that they’re only looking for volunteers right now). A product designer with an AI safety portfolio, applying for a product design role within an AI safety company. Listen, I’m not saying I was entitled to the job, but there aren’t exactly swarms of people who fit the bill here. I figured I’d at the very least get an interview.
Nope. Automated rejection. So, I did what any self-respecting cog-in-the-machine does and went to the team page on their website, hoping to figure out who got the job over me so I could stalk their LinkedIn profile and feel even shittier about myself. I couldn’t find them, but I did find something way worse (which, ironically, made me feel slightly better in the moment): excluding the CEO and COO, they list 31 technical employees, two of whom are women. TWO!
And the org with the graphic design request isn’t all that much better, just smaller. They currently list 12 fellows for a program I was turned away from, and only two are women. Two out of 12 beats two out of 31, but neither even comes close to EA’s own gender distribution (71% male, per this survey), let alone gender parity, and that is a low bar.
Oooh, sexism in tech. Groundbreaking, I know.
But then again, this isn’t just tech, is it? It’s AI safety, a field that exists to ensure evolving technologies align with human values, not our worst biases. Yeah, good luck keeping those biases out of the algorithms when you can’t even keep them out of your hiring practices.
Okay, so maybe I am a little butthurt over the rejections. You would be too if you watched, in real time, people — men — who started at the same place as you get handed opportunities on promise alone, while your prototype couldn’t even land you an interview.
There’s a name for this, I learned. The prove-it-again bias: men are judged on potential, while women are judged on what they’ve already delivered, and then asked to deliver it again. I shipped a product that funders liked and effectively got told to piss off but stay on hold in case they needed a volunteer, while men who haven’t even graduated college yet — no industry experience to their name — effortlessly walked through the doors.
And the worst part of it all is that even now — even after being dragged through the mud by this movement — my first instinct is still to say yes to an org that didn’t want me but wanted my free labor. Because what I’m after is not a paycheck — I already have one of those, and it’s solid enough that animal welfare orgs get to see 10% of it every month. What I’m after is the opportunity to help as many animals, current and future, as my skills will allow.
Which is probably exactly where this movement wants me. “Set your ego aside.” “The mission is bigger than you.” Saying no here feels like I’m choosing myself over animals, because the cost of the closed loop isn’t my dignity; it’s potentially billions of their lives.
So, I’m still putting in the work. Still upskilling, still learning, and still building — but I’ve started weaving in some impact calculations of my own to make sure my time goes where it’ll count. And per my calculations, the expected value of repeatedly throwing myself at a closed loop is… approximately zero. Which must mean, then, that the maximum value I can create for animals has to be outside of the loop. And so from here on out, that is exactly where my ideas and I will live.
But don’t count on my math. After all, I don’t have a technical background.