Aaron Bergman

Karma: 1,706

I graduated from Georgetown University in December, 2021 with degrees in economics, mathematics and a philosophy minor. There, I founded and helped to lead Georgetown Effective Altruism. Over the last few years recent years, I’ve interned at the Department of the Interior, the Federal Deposit Insurance Corporation, and Nonlinear, a newish longtermist EA org.

I’m now doing research thanks to an EA funds grant, trying to answer hard, important EA-relevant questions. My first big project (in addition to everything listed here) was helping to generate this team Red Teaming post.

Blog: aaronbergman.net

Aaron Bergman 22 Apr 2024 13:13 UTC
1 point
0 ∶ 2
in reply to: Ben Millwood’s comment on: BenMillwood’s Shortform
Automated interface between Twitter and the Forum (eg a bot that, when tagged on twitter, posts the text and image of a tweet on Quick Takes and vice versa)

Aaron Bergman 17 Apr 2024 9:20 UTC
58 points
14 ∶ 1
on: How good it is to donate and how hard it is to get a job
I’d be very surprised if you can’t get a job that pays much more than the sub teacher role- the gap between that and ~any EA org job is massive and inability to get the latter is only very weak evidence of inability to earn more.

Sorry if I missed this but this does depend a lot on location/willingness to move. The above assumes If you’re in the US and willing to move cities.

Also, living frugally to donate more is of course very virtuous if you take your salary to be a given, but from an altruistic perspective, insofar as they trade off, it’s probably much better to spend effort on finding a way to earn more.

This is a bit of a hobbyhorse of mine but this could look like “found a startup with a 5% chance of earning $10M” in addition to or instead of searching for higher salaried roles

Aaron Bergman 15 Mar 2024 0:14 UTC
3 points
3 ∶ 0
on: AIM (CE) new program: Founding to Give. Apply now to launch a high-growth company!
Hell yeah this is awesome

Aaron Bergman 10 Mar 2024 22:51 UTC
6 points
0 ∶ 0
in reply to: Will Howard’s comment on: Podcast idea: interview low-key people doing direct work
Omg 😊😊😊😊

Aaron Bergman 9 Mar 2024 12:57 UTC
11 points
1 ∶ 2
on: aaronb50′s Shortform
Random sorta gimmicky AI safety community building idea: tabling at universities but with a couple laptop signed into Claude Pro with different accounts. Encourage students (and profs) to try giving it some hard question from eg a problem set and see how it performs. Ideally have a big monitor for onlookers to easily see.

Most college students are probably still using ChatGPT-3.5, if they use LLMs at all. There’s a big delta now between that and the frontier.

Aaron Bergman 13 Jan 2024 5:26 UTC
27 points
1 ∶ 0
on: aaronb50′s Shortform
I made a custom GPT that is just normal, fully functional ChatGPT-4, but I will donate any revenue this generates^[1] to effective charities.
Presenting: Donation Printer
1. ^
  OpenAI is rolling out monetization for custom GPTs:
  Builders can earn based on GPT usage
  In Q1 we will launch a GPT builder revenue program. As a first step, US builders will be paid based on user engagement with their GPTs. We’ll provide details on the criteria for payments as we get closer.

Aaron Bergman 1 Jan 2024 12:07 UTC
5 points
0 ∶ 0
in reply to: Charles Dillon ’s comment on: aaronb50′s Shortform
Yeah you’re right, not sure what I missed on the first read

Aaron Bergman 1 Jan 2024 4:00 UTC
2 points
0 ∶ 0
in reply to: Charles Dillon ’s comment on: aaronb50′s Shortform
This doesn’t obviously point in the direction of relatively and absolutely fewer small grants, though. Like naively it would shrink and/or shift the distribution to the left—not reshape it.

Aaron Bergman 31 Dec 2023 2:01 UTC
8 points
2 ∶ 0
in reply to: Charles Dillon ’s comment on: aaronb50′s Shortform
Yeah but my (implicit, should have made explicit lol) question is “why this is the case?”

Like at a high level it’s not obvious that animal welfare as a cause/field should make less use of smaller projects than the others. I can imagine structural explanations (eg older field → organizations are better developed) but they’d all be post hoc.

Aaron Bergman 30 Dec 2023 21:47 UTC

28 points

2 ∶ 0

on: aaronb50′s Shortform

Interesting that the Animal Welfare Fund gives out so few small grants relative to the Infrastructure and Long Term Future funds (Global Health and Development has only given out 20 grants, all very large, so seems to be a more fundamentally different type of thing(?)). Data here.

A few stats:

The 25th percentile AWF grant was $24,250, compared to $5,802 for Infrastructure and $7,700 for LTFF (and median looks basically the same).
AWF has only made just nine grants of less than $10k, compared to 163 (Infrastructure) and 132 (LTFF).

Proportions under $threshold

fund	prop_under_1k	prop_under_2500	prop_under_5k	prop_under_10k
Animal Welfare Fund	0.000	0.004	0.012	0.036
EA Infrastructure Fund	0.020	0.086	0.194	0.359
Global Health and Development Fund	0.000	0.000	0.000	0.000
Long-Term Future Fund	0.007	0.068	0.163	0.308

Grants under $threshold

fund	n	under_2500	under_5k	under_10k	under_25k	under_50k
Animal Welfare Fund	250	1	3	9	243	248
EA Infrastructure Fund	454	39	88	163	440	453
Global Health and Development Fund	20	0	0	0	5	7
Long-Term Future Fund	429	29	70	132	419	429

Summary stats (rounded)

fund	n	median	mean	q1	q3	total
Animal Welfare Fund	250	$50,000	$62,188	$24,250	$76,000	$15,546,957
EA Infrastructure Fund	454	$15,319	$41,331	$5,802	$45,000	$18,764,097
Global Health and Development Fund	20	$900,000	$1,257,005	$297,925	$1,481,630	$25,140,099
Long-Term Future Fund	429	$23,544	$44,624	$7,700	$52,000	$19,143,527

Aaron Bergman 24 Dec 2023 20:06 UTC
12 points
0 ∶ 0
on: On the future of language models
In their most straightforward form (“foundation models”), language models are a technology which naturally scales to something in the vicinity of human-level (because it’s about emulating human outputs), not one that naturally shoots way past human-level performance
- i.e. it is a mistake-in-principle to imagine projecting out the GPT-2—GPT-3—GPT-4 capability trend into the far-superhuman range
Surprised to see no pushback on this yet. I do not think this is true; I’ve come around to thinking that Eliezer is basically right that the limit of next token prediction on human generated text is superintelligence. Now how this latent ability manifests is a hard question, but it’s there to be used by the model for its own ends or elicited by humans for ours, or both.
Also worth adding (guessing this point has been made before) that non human-generated text (e.g. regression outputs from a program) are in the training data, so merely predicting those gets you superhuman performance in some domains.

Aaron Bergman 21 Nov 2023 18:57 UTC
32 points
10 ∶ 7
on: Rethink Priorities needs your support. Here’s what we’d do with it.
For others considering whether/where to donate: RP is my current best guess of “single best charity to donate to all things considered (on the margin—say up to $1M).”
FWIW I have a manifold market for this (which is just one source of evidence—not something I purely defer to. Also I bet in the market so grain of salt etc).

Aaron Bergman 19 Nov 2023 19:19 UTC
89 points
16 ∶ 2
on: Open Phil Should Allocate Most Neartermist Funding to Animal Welfare
Strongly, strongly, strongly agree. I was in the process of writing essentially this exact post, but am very glad someone else got to it first. The more I thought about it and researched, the more it seemed like convincingly making this case would probably be the most important thing I would ever have done. Kudos to you.
A few points to add
1. Under standard EA “on the margin” reasoning, this shouldn’t really matter, but I analyzed OP’s grants data and found that human GCR has been consistently funded 6-7x more than animal welfare (here’s my tweet thread this is from)
2. @Laura Duffy’s (for Rethink Priorities) recently published risk aversion analysis basically does a lot of the heavy lifting here (bolding mine):
Spending on corporate cage-free campaigns for egg-laying hens is robustly^[8] cost-effective under nearly all reasonable types and levels of risk aversion considered here.
1. Using welfare ranges based roughly on Rethink Priorities’ results, spending on corporate cage-free campaigns averts over an order of magnitude more suffering than the most robust global health and development intervention, Against Malaria Foundation. This result holds for almost any level of risk aversion and under any model of risk aversion.
I also want to emphasize this part, because it’s the kind of serious engagement with suffering that EA still fails to to do enough of
I experienced “disabling”-level pain for a couple of hours, by choice and with the freedom to stop whenever I want. This was a horrible experience that made everything else seem to not matter at all.
A single laying hen experiences hundreds of hours of this level of pain during their lifespan, which lasts perhaps a year and a half—and there are as many laying hens alive at any one time as there are humans. How would I feel if every single human were experiencing hundreds of hours of disabling pain?
A single broiler chicken experiences fifty hours of this level of pain during their lifespan, which lasts 4-6 weeks. There are 69 billion broilers slaughtered each year. That is so many hours of pain that if you divided those hours among humanity, each human would experience about 400 hours (2.5 weeks) of disabling pain every year. Can you imagine if instead of getting, say, your regular fortnight vacation from work or study, you experienced disabling-level pain for a whole 2.5 weeks? And if every human on the planet—me, you, my friends and family and colleagues and the people living in every single country—had that same experience every year? How hard would I work in order to avert suffering that urgent?
Every single one of those chickens are experiencing pain as awful and all-consuming as I did for tens or hundreds of hours, without choice or the freedom to stop. They are also experiencing often minutes of ‘excruciating’-level pain, which is an intensity that I literally cannot imagine. Billions upon billions of animals. The numbers would be even more immense if you consider farmed fish, or farmed shrimp, or farmed insects, or wild animals.
If there were a political regime or law responsible for this level of pain—which indeed there is—how hard would I work to overturn it? Surely that would tower well above my other priorities (equality, democracy, freedom, self-expression, and so on), which seem trivial and even borderline ridiculous in comparison.

Aaron Bergman 14 Oct 2023 5:32 UTC
10 points
1 ∶ 4
on: If Contractualism, Then AMF
[On mobile; sorry for the formatting]

Given my quick read and especially the bit below, it seems like the title is at least a bit misleading.

Quote: “To be clear: this document is not a detailed vindication of any particular class of philanthropic interventions. For example, although we think that contractualism supports a sunnier view of helping the global poor than funding x-risk projects, contractualism does not, for all our argument implies, entail that many EA-funded global poverty interventions are morally preferable to all other options (some of which are probably high-risk, high-reward longshots).”

I think a reasonable person would conclude from the title “If Contractualism, Then AMF” essentially the opposite of this more nuanced clarification.

Perhaps it’s reasonable to infer that “Then AMF” really means “then the cluster of beliefs that leads GiveWell to strongly recommend AMF are indeed true (even if ex post it turns out that deworming or something was better)” but even this doesn’t seem to be what you are arguing (given the quote above).

Aaron Bergman 12 Oct 2023 2:46 UTC
6 points
1 ∶ 0
on: aaronb50′s Shortform
LessWrong has a new feature/type of post called “Dialogues”. I’m pretty excited to use it, and hope that if it seems usable, reader friendly, and generally good the EA Forum will eventually adopt it as well.

Aaron Bergman 9 Oct 2023 22:23 UTC
8 points
1 ∶ 1
on: Potentially actionable opportunity: eliminating the New World screwworm (flesh-devouring maggots that affect a billion animals each year)
I’m interested in supporting this financially (that sounds like something a rich person would say so I should clarify this would not be a ton of money lol) and possibly in other ways as well (e.g., helping set up a website)

Aaron Bergman 9 Oct 2023 22:23 UTC
2 points
0 ∶ 0
in reply to: Yellow (Daryl)’s comment on: Potentially actionable opportunity: eliminating the New World screwworm (flesh-devouring maggots that affect a billion animals each year)
At least some chance of a less terrible death later, no? I’m really not sure what the distribution of causes of death looks like for different types of wild animal hosts

Aaron Bergman 9 Oct 2023 7:05 UTC
4 points
0 ∶ 0
on: aaronb50′s Shortform
New fish data with estimated individuals killed per country/year/species (super unreliable, read below if you’re gonna use!)
That^ is too big for Google Sheets, so here’s the same thing just without a breakdown by country that you should be able to open easily if you want to take a look.
Basically the UN data generally used for tracking/analyzing the amount of fish and other marine life captured/farmed and killed only tracks the total weight captured for a given country-year-species (or group of species).
I had chatGPT-4 provide estimated lower and upper bounds on the average weight of individual members of given species/groups, thereby allowing me to guesstimate individual numbers of fish+ from the UN live weight data. I only did this for species/groups of which >5000 metric tons have been harvested since 1961, which covers about 99.9% of the total mass.
Happy to describe anything in more detail if it would actually be of use!

Aaron Bergman 8 Oct 2023 23:15 UTC
15 points
6 ∶ 1
in reply to: Jonathan Paulson’s comment on: Our planned allocation to GiveWell’s recommendations for the next few years
Good point, and I’ll throw out The Humane League as one specific recipient of money.
Farmed animal welfare is politically controversial in a way that GiveWell is not. This is potentially bad:
Is OpenPhil’s current support of farmed animal welfare politically controversial? I don’t get that sense but, if so, among who?
Maybe people who don’t care about farmed animals are correct
Sure but same goes for literally everything, including eg AMF being net positive. Happy to discuss object level though.
Farmed animal advocacy is so cost-effective because, if successful, it forces other people (meat consumers? meat producers?) to bear the costs of treating animals better. I’m less comfortable spending other people’s money to make the world better than spending my own money to make the world better
Interesting point and yeah I think this is valid. At some margin I think this would become an important consideration (e.g., advocating some policy that made being non-vegan super expensive) but at the current margin it seems like these costs are just extremely small relative to the suffering reduction they induce.
Increased advocacy for farm animals might just cause increased advocacy for farms, just burning money rather than improving the world
Farm lobby is strong. I agree this has to be accounted for but trust OpenPhil, ACE, and e.g. The Humane League to account for this when deciding what to do and who to fund. Empirically, it seems to be the case that e.g. cage free advocacy has worked and laws like California’s prop 12 have passed and been upheld.
It’s hard to be as confident in political interventions—humans and groups of humans are much less predictable than e.g. malaria
First, at one level I agree but then would point to all the non-political animal welfare interventions like cage free advocacy without bans and Shrimp Welfare Project paying for farms to install stunners. At another I just disagree that e.g. AMF has high-confidence certain impact on the world. All the analyses explicitly don’t even try to account for 3rd+ order effects (not sure about 2nd) which is plausibly where a ton of impact lies.
- Farmed animal welfare sometimes seems overly-connected with dubious left-wing politics (e.g. https://forum.effectivealtruism.org/posts/5iCsbrSqLyrfP55ry/concerns-with-ace-s-recent-behavior-1)
I’ve definitely seen first hand how much of especially veganism/vegan advocacy per se is very lefty and, more importantly, less clear eyed and epistemically rigorous than EA in general and certainly meta level EA orgs. IMO the appropriate response here is to be a countervailing force in the sense of technocratic rigor (not conservatism), not to “leave those people be”.

Aaron Bergman 7 Oct 2023 6:29 UTC
25 points
26 ∶ 41
on: Our planned allocation to GiveWell’s recommendations for the next few years
I’ve argued this largely on Twitter, but it seems pretty clear to me that no marginal dollars at all, at least up to say $1B, should in fact be going to the GiveWell portfolio (or similar charities for that matter). I don’t think it’s obvious what the alternative should be, but do think that (virtually) no well informed person trying to allocate a marginal dollar most ethically would conclude that GiveWell is the best option.

I feel like this/adjacent debates often gets framed as “normal poverty stuff vs weird longtermist stuff” but a lot of my confidence in the above comes from farmed animal welfare strictly dominating GiveWell in terms of any plausibly relevant criteria save for maybe PR. And then there’s an important and interesting debate to be had over farmed animals vs GCR vs everything else.

I’d be super keen to hear from anyone who disagrees from an affirmative “no really, AMF etc is more deserving than every other org I know about” perspective, as I don’t think I’ve encountered anyone who’s argued this.

Aaron Bergman

Builders can earn based on GPT usage

A few points to add