Aspiring EA from Netherlands (Indian by birth)
agent18
Whoever downvoted this, would you care to inform me why? Is the data not accurate? Am I hating on 80k too much? what is it?
PDF link doesn’t exist anymore. @Peter_Hurford
@Carrickflynn,
It’s been 3 years now. Is it possible to do a retake evaluating the current situation of the Disentanglement and if there has been growth in possibilities to work in AI strategy & policy (be in implementation or research)?
Thanks.
Nice Post Aaron! I have the following questions:
1. I was wondering if you can also provide the score ruberic and distribution for the interview rounds and the final work trial rounds?
2. Within what time frame did you receive the 180+ applications?
Thank you very much Aaron. Are you then able to inform the distribution of the scores for the interview (21 people) and the final work trail (8 people)? I understand they are subjective. Nevertheless they were a score on 10.
And maybe this is a bit much--> Do you have the distribution of where you got your candidates from? Here is an example from EAF’s hiring round
Hi Aaron, Can you also answer the following for me please?
So, in addition to hiring a candidate, I’ve also kept a record of the other applicants who most >impressed me, so that I can let them know if I hear about promising opportunities. I’ve already >referred a few candidates for different part-time roles at other EA orgs, and I anticipate more >chances to come.
-
How many people “most impressed you”?
-
How many people have you already referred for different part-time roles at other EA orgs?
-
How many people do you think EA orgs are hiring in this job type currently or within the last year?
-
How many people in your list who didn’t get hired, do you expect to get hired else where in EA orgs? (gut feel, guess, based on past experiences, anything)
-
Thank You for your post. I have placed quotes on things that I am not sure I understand you correctly. Hence I seek an example.
I was wondering if you have one example for this “median EA” who got into an “early stage project” as a result of getting into the “EA network”. And can you also inform how the example-EA got into the “EA network”?
This is still a bit rate limited, and couldn’t handle an influx of thousands of people. But I think it can handle more than it currently does.
Why do you say it is rate limited and that it can handle more “median EAs” than it currently does? I hope you can give an example for this or perhaps quote your experience if you are up close to such events.
[Question] Examples for impact of Working at EAorg instead of ETG
Very much appreciate the detailed response. I think you have answered both my questions. Very much appreciate the clear example. If there are only 100 jobs in EA per year, it seems unlikely to support 1000s in the way you have suggested (rate limited).
How does a “median EA” look?
he (the median EA) is within the 60-90th percentile (I am unsure of what, IQ?)
In the case with LW, he was able to talk about rationality and the “surrounding ecosystem”. If you can, I would really like an example for this?
P.S I am trying to judge if I could be a potential “median-EA”. Hence the questions.
Thanks.
A naive analysis on if EA is Talent constrained
Do you know of actual TC positions? Can you please cite your source?
If you find formatting issues please state here:
Claims that you find to be false? please post evidence as well.
I corrected it. Thanks.
Very interesting! Can you please cross-post?
Thank You for acknowledging this post. I very much appreciate your reply.
We’ve been trying to do a better job communicating our uncertainty in the new key ideas series, for instance by releasing: advice on how to read our advice
I really wish you can put more of your evidence out there instead of sentences that are a summary of the evidence you have. “Another bottleneck to progress on GPR might be operations staff” (GPR Key-ideas). Is it a bottleneck or is it not? I don’t know what to make of “might be”. In this case if you presented your evidence that helps conclude this, say in a footnote, I think it will be more useful. People can then draw the conclusion for themselves.
To be specific, I think it’s longtermist organisations that are most talent constrained. Global health and factory farming organisations are much more constrained by funding relatively speaking (e.g. GiveWell top recommended charities could absorb ~$100m). I think this explains why organisations like TLYCS, Charity Science and Charity Entrepreneurship say they’re more funding constrained (and also to some extent Rethink priorities, which does a significant fraction of its work in this area).
I am glad you clarify about your position that you are focused on longtermism TC. I only know of two cases where longtermism positions are TC. Disentanglement research as informed by Carrick Flynn in Sep 2017 and AI Policy in US in Jan 2019 article). It still stands that Open Phil in GR seems to be not TC. (“The pool of available talent is strong, … more than a hundred applicants had very strong resumes… but … (to) deploy this base of available talent is weak”)
I think what helps is to keep the TC debate focused on to specific cases. And this can be done with providing evidence as done in AI Policy in US.
Even within longtermist and meta organisations, not every organisation is mainly skill-constrained, so you can find counterexamples, such as new organisations without much funding. This may also explain the difference between the average survey respondents and Rethink Priorities’ view.
Claims: Average Survey respondents feel they are TC more than RP because they have less funding needs than RP (and is “new”).
Example: Open Phil is an average survey respondent (I presume). Open Phil has funding. Open Phil does not seem to feel TC in GR though.
It looks like the example does not satisfy the claim. So now I don’t really know what you are talking about. I don’t have one example of an org and a position that is skill-constrained in research in GPR. I keep hearing you saying that “research is the biggest need right now” (key-ideas post) but when I look in Open Phil it doesn’t seem to be so. They are unable to absorb more researchers. So what exactly are you talking about?
You might wonder why I am quoting the same Open Phil example like a parrot. That is because that is one of the few hiring rounds available. And trying to ask companies like FHI or Open Phil etc., for more info on this or dollars moved by researcher or about replaceability does not seem to produce results unfortunately.
It doesn’t seem to me that looking at whether lots of people applied to a job tells us much about how talent constrained an organization is.
The definition for TC is that an org is unable to find “skilled people” despite hiring actively. I agree that number of people applied is not a measure for TC. But the number of people in the last round (after 4 other rounds) seems to suggest something regarding if orgs are able to find skilled people or not. Even if that is not the case --> When you look at what Open Phil says, I can’t imagine that they are TC in GR based on the numbers of people who they thought had good resumes. In fact it seems like a bad idea to push for research at Open Phil (GPR) in GR considering replaceability atleast. And the more I talk to people like Peter Hurford (about replaceability) the more I feel like there is less point in being a GR.
Some successful applicants might have still been much better than others, or the organisations might have preferred to hire even more than they were able to.
About “successful applicants might have been still much better” (due to the potential log-normal distribution of candidates ability), I would also like one example for a case where this is true. I don’t think that is the case with Open Phil in GR based on their hiring round.
Aaron also raised this point as well. Yes that is definitely a possibility that people would still be hired but the organization would continue to be TC. Seems like a reasonable hypothesis but still needs evidence (one example at least) to support it I think. Nevertheless, I don’t think that is the case with Open Phil in GR based on their hiring round.
Something else I think is relevant to the question of whether our top problem areas are talent constrained is that I think many community members should seek positions in government, academia and other existing institutions. These roles are all ‘talent constrained’, in the sense that hundreds of people could take these positions without the community needing to gain any additional funding. In particular, we think there is room for a significant number of people to take AI policy careers, as argued here.
AI policy careers in the US seems to match the definition of TC. “80,000 Hours has attended, speakers have lamented the government’s lack of expertise on AI, and noted the substantial demand for such expertise within government. For example, DoD’s new Joint AI Center alone is apparently looking to hire up to 200 people.”. I didn’t know this before. This is so clear for me now, that I have an example for what you mean with “significant number of people”. I wish the same was available for other top problem areas.
Thanks for this.
Thank you very much for taking the time to respond. I very much appreciate it. I would really appreciate more evidence displayed for claims and less generalization with 80khours blogs.
P.S
If you already know many opportunities are high-impact, I expect that you have looked at the value contributed by several people, and factored things like replaceability etc., before you came to a decision. Why not just publish it? Asking companies doesn’t seem practical and no one seems to be giving out such information. One author even suggested that only if I am writing an academic paper he would be able to help otherwise he didn’t find time for it.
Hi Jamie,
Thank You for your comment.
Isn’t TC in the movement just the aggregation of TC in relevant orgs and actors?
Yes it seems to be. All I wanted was to avoid a level of abstraction. “AI strategy is TC in DR” vs “FHI is TC in DR”. I really feel confused thinking about the former. The later is so concrete. I can test it. I can go in depth in that ONE EXAMPLE. The former is too broad. I find it easier to think in concrete examples.
There’s a tradeoff between specificity/concreteness and representatives/unreliability, and for most purposes, the latter seems more useful to me?
Interesting! Would you be able to give me a real example to satisfy your claim? I claim that concreteness seems useful to me and if I get an example I hold on to it for dear life and test all claims atleast against that one example.
Claim: Concreteness seems useful.
Example: Consider: “Many community members should seek positions in government, academia, and other existing institutions.”
I am lost. What is “MANY”? What does a “position in government” even look like. All this until I saw this beautiful example: “DoD’s new Joint AI Center alone is apparently looking to hire up to 200 people.”. I understand finally what many and position in government is.
Animal Advocacy Careers will be offering one-to-one advising soon. Before it is officially launched, people can sign up to express their interest here.
That’s great. I subscribed already. Thank You very much Jamie.
It appears you are extremely good in your field (16th out of 6000). A heart surgeon in the US has a median salary of 448k dollars. As much as 80k claims that being a doctor is pointless, Imagine the ETG as a result of this. This is close to avg expected salaries of top traders (600k dollars) in the game.