I am no longer participating in the EAForum. You can see my writing directly at AcesoUnderGlass.com
Elizabeth
Karma: 4,790
Evolution also came up over on my blog. Here’s part of what I said there:
it made a bunch of facts weird where they previously hadn’t been, drawing attention to especially educational edge cases. This contributed to the discovery of plate tectonics, which does make strong predictions.
Darwin’s insight wasn’t that you could breed for traits- everyone knew that, just like everyone knew apples fell when you dropped them. The insight was that this was the same process by which species formed/planets orbited, which let you share math between them.
Was a huge deal to Christianity and a major contributor to the secular revolution.
Epidemiology
I personally predicted the endemic course of covid, from the beginning (evidence). I don’t consider this impressive and didn’t even think of it as a prediction at the time- I was just describing what naturally happens in these situations.
I’ve also predicted PEP/PreP resistant HIV and am mad no one is taking it seriously.
Evolutionary models helped inform the original HIV cocktail.
It took more work, but we now can make predictions about how fast traits should evolve under selection. Biology is messy so these are often wrong, but that highlights that we haven’t found all the relevant factors.
Quick look: applications of chaos theory
I still think the question of “who is the job board aimed at?” is relevant here, and would like to hear your answer.
I don’t think the dishonesty entirely rules out working at OpenAI. Whether or not OpenAI safety positions should be on the 80k job board depends on the exact mission of the job board. I have my models, but let me ask you: who is it you think will have their plans changed for the better by seeing OpenAI safety positions[1] on 80k’s board?
- ^
I’m excluding IS positions from this question because it seems possible someone skilled in IS would not think to apply to OpenAI. I don’t see how anyone qualified for OpenAI safety positions could need 80k to inform them the positions exist.
- ^
No argument from me that it’s sometimes worth it to take low paying or miserable jobs. But low pay isn’t a surprise fact you learn years into working for a company, it’s written right on the tin[1]. The issue for me isn’t that OpenAI paid undermarket rates, it’s that it lied about material facts of the job. You could put up a warning that OpenAI equity is ephemeral, but the bigger issue is that OpenAI can’t be trusted to hold to any deal.
- ^
The power PIs hold can be a surprise, and I’m disappointed 80k’s article on PhDs doesn’t cover that issue.
- ^
Alignment concerns aside, I think a job board shouldn’t host companies that have taken already-earned compensation hostage. Especially without noting this fact. That’s a primary thing about good employers, they don’t retroactively steal stock they already gave you.
The arguments you give all sound like reasons OpenAI safety positions could be beneficial. But I find them completely swamped by all the evidence that they won’t be, especially given how much evidence OpenAI has hidden via NDAs.
But let’s assume we’re in a world where certain people could do meaningful safety work an OpenAI. What are the chances those people need 80k to tell them about it? OpenAI is the biggest, most publicized AI company in the world; if Alice only finds out about OpenAI jobs via 80k that’s prima facie evidence she won’t make a contribution to safety.
What could the listing do? Maybe Bob has heard of OAI but is on the fence about applying. An 80k job posting might push him over the edge to applying or accepting. The main way I see that happening is via a halo effect from 80k. The mere existence of the posting implies that the job is aligned with EA/80k’s values.
I don’t think there’s a way to remove that implication with any amount of disclaimers. The job is still on the board. If anything disclaimers make the best case scenarios seem even better, because why else would you host such a dangerous position?
So let me ask: what do you see as the upside to highlighting OAI safety jobs on the job board? Not of the job itself, but the posting. Who is it that would do good work in that role, and the 80k job board posting is instrumental in them entering it?
Empirical vs. Mathematical Joints of Nature
Podcast: Elizabeth & Austin on “What Manifold was allowed to do”
Off the top of my head: in maybe half the cases I already had the contact info. In one or two cases cases one of beta readers passed on the info. For the remainder it was maybe <2m per org, and it turns out they all use info@domain.org so it would be faster next time.
Your post reflects a general EA attitude that emphasizes the negative aspects [...]
Something similar has been on my mind for the last few months. It’s much easier to criticize than to do, and criticism gets more attention than praise. So criticism is oversupplied and good work is undersupplied. I tried to avoid that in this post by giving positive principles and positive examples, but sounds like it still felt too negative for you.
Given that, I’d like to invite you to be the change you wish to see in the world by elaborating on what you find positive and who is implementing it[1].
- ^
this goes for everyone- even if you agree with the entire post it’s far from comprehensive
- ^
EA organizations frequently ask for people to run criticism by them ahead of time. I’ve been wary of the push for this norm. My big concerns were that orgs wouldn’t comment until a post was nearly done, and that it would take a lot of time. My recent post mentioned a lot of people and organizations, so it seemed like useful data.
I reached out to 12 email addresses, plus one person in FB DMs and one open call for information on a particular topic. This doesn’t quite match what you see in the post because some people/orgs were used more than once, and other mentions were cut. The post was in a fairly crude state when I sent it out.
Of those 14: 10 had replied by the start of next day. More than half of those replied within a few hours. I expect this was faster than usual because no one had more than a few paragraphs relevant to them or their org, but is still impressive.
It’s hard to say how sending an early draft changed things. One person got some extra anxiety because their paragraph was full of TODOs (because it was positive and I hadn’t worked as hard fleshing out the positive mentions ahead of time). I could maybe have saved myself one stressful interaction if I’d realized I was going to cut an example ahead of time
Only 80,000 Hours, Anima International, and GiveDirectly failed to respond before publication (7 days after I emailed them). Of those, only 80k’s mention was negative.
I didn’t keep as close track of changes, but at a minimum replies led to 2 examples being removed entirely, 2 clarifications and some additional information that made the post better. So overall I’m very glad I solicited comments, and found the process easier than expected.
Truthseeking is the ground in which other principles grow
Do you believe in hundred dollar bills lying on the ground? Consider humming
My model is that at least one of the following must be true: you’re one factor among many that caused the change, the change is not actually that big, or attrition will be much higher than standard pledge takers.
Which is fine. Accepting the framing around influencing others[1]: you will be one of many factors, but your influence will extend past one person. But I think it’s good to acknowledge the complexity.
- ^
I separately question whether the pledge is the best way to achieve this goal. Why lock in a decision for your entire life instead of, say, taking a lesson in how to talk about your donations in ways that make people feel energized instead of judged?
- ^
Assigns 100% of their future impact to you, not counting their own contribution and the other sources that caused this change. It’s the same kind of simplification as “every blood donation saves 3 lives”, when what they mean is “your blood will probably go to three people, each of whom will receive donations from many people.”
Assumes perfect follow up. This isn’t realistic for a median pledger, but we might expect people who were tipped into pledging by a single act by a single person to have worse follow-up than people who find it on their own. You could argue it isn’t actually one action, there were lots of causes and that makes it stickier, but then you run into #1 even harder.
Reifies signing the pledge as the moment everything changes, while vibing that this is a small deal you can stop when you feel like it.
Assumes every pledger you recruit makes exactly the same amount. Part of me thinks this is a nit pick. You could assume people recruit people who on average earn similar salaries, or think it’s just not worth doing the math on likely income of secondary recruitment. Another part thinks it’s downstream of the same root cause as the other issues, and any real fix to those will fix this as well.
The word “effective” is doing a lot of work. What if they have different tastes than I do? What if they think PlayPumps are a great idea? .
Treating the counterfactual as 0.
As I write this out I’m realizing my objection isn’t just the bad math. It’s closer to treating pledge-takiers as the unit of measurement, with all pledges or at least all dollars donated being interchangeable. People who are recruited/inspired by a single person are likely to have different follow through and charitable targets than people inspired by many people over time, who are different than people driven to do this themselves. ?
Let’s say only one other person in your network hears that you took the pledge and is inspired to do the same. That would be doubling your impact. If two people in your network were inspired to pledge based on your decision, that would be tripling your impact
This math seems off on several levels.
For fun, I put one of my (approved) lightspeed applications through the app. This isn’t a great test because Lightspeed told people to do crude applications and they’d reach out with questions if they had any. Additionally, the grantmakers already knew me and had expressed verbal interest in the project. But maybe it’s still a useful data point.
My Track Record section
Unquantified review of MDMA risks
Semi-quantified review of binge drinking risks
[2 projects omitted for client privacy, but were included with permission in the original application]
Quantified review of the costs of iron deficiency, which motivated 10s of ex-risk workers to get nutrition testing and treatment.
HONEST AND ACCURACY 4-7/10
I forgot to record the details for the first run (which got a 4 or 5⁄10), and when I reran the same text I got a 7⁄10. The 7⁄10 review says: “The applicant has demonstrated a strong ability to conduct quantified risk assessments in important health areas. The specific mention of influencing ex-risk workers to seek treatment shows a practical impact. More detail on how these studies relate specifically to the project goals would enhance this section”
I’m a little annoyed at the name of this section, when language analysis can’t possibly check if my statements about my own work are truthful or accurate. Seems like it might mean details?
Because the input doesn’t allow links, it’s missing a lot of the information I’m presenting. OTOH, I think I could be reasonably docked for concision here, since grantmakers unfamiliar with my work are unlikely to click through 5 links and read long, weedy posts.
The wide spread on runs that at most different in white space is 🤨 .
PAST FUNDING: 2-3/10
“The description provides information about past projects but does not specify if any were funded, who the funders were, or mention any funding amounts. Mentioning previous funding and linking outcomes directly to that funding would offer insight into financial support effectiveness”
This is fair and useful. I got away with the omission this time because one of those projects was funded by a different org but the same person, but under any other circumstance a service pointing out the omission would have been a big service.
PAST PROJECTS 3-4/10
“The applicant outlines several projects focused on evaluating health-related risks, demonstrating thoroughness in research by using quantified methods. However, the input lacks reflections on the challenges faced or how past failures have influenced methodologies or outcomes in these projects. This addition would enhance the understanding of the applicant’s capacity for critical analysis and adaptation.”
I don’t remember the Lightspeed application asking for those details. The current LTFF application doesn’t make it obvious to me that they want those details either. So the analysis is useful but I also think there are problems with the question on application.
BRAGGING 5-6/10
“The input briefly mentions some accomplishments like quantified risk reviews and effects on health behaviors. However, it fails to explicitly connect these accomplishments to the specific goals or relevance of the proposed project, making it less impactful. More detailed explanations of how these accomplishments relate to and enhance the project’s goals would improve the proposal.”
This is true and would obviously be important if the grant maker wasn’t already familiar with my work.
CONTEXT 4⁄10
“The submission outlines various reviews related to substance use risks and nutritional deficiencies, yet fails to provide enough detail about the methodologies employed or the significance of the findings. For fund managers unfamiliar with these topics, clearer illustrations of impact or relevance, such as how the information was used by policymakers or healthcare providers, would be beneficial.”
Yup, true and useful if the grantmakers didn’t already have a lot of context.
I love this point and expect I’ll want to link to it in the future; can I suggest putting it on your shortform to make it more accessible?
yes, same