Here’s our update since April, summarising our progress over bout 2 and our plans for bout 3. See our previous update.
Progress in two sentences
We focused on (i) releasing articles (ii) deciding the vision for advising (iii) testing out ‘person-first’ headhunting (iv) research. Habiba Islam started on the advising team (having accepted the offer last year) and got up-to-speed quickly, and Luisa Rodriguez accepted a job offer to start in 2021 as a researcher – more info below.
We closed our Dec 2019 round of $3.9m several months ago.
On the 11 July, we also closed the initial target for the Dec 2020 round of $4.1m, which covers us for 2021 while adding 1.5 FTE.
This December, we expect to aim to raise around $5m in commitments towards our Dec 2021 round and topping up Dec 2020, and so any future donations we receive are put towards that target.
See more information on our annual review and fundraising targets here.
Note that our cash reserves are unusually high because we received some donations as cash that would have normally been future commitments, and so far we’ve underspent our expansion budget in 2020.
-----------------------Full update -----------------------
Story of a plan change
In 2017 Lewis Ho was an economics major at Yale University intending to become a development economist. In August of that year, he applied for one-on-one advising, wondering whether he should stick to development economics or switch to global priorities research or AI.
During our call, we encouraged him to explore these areas and sketched out a plan for him to investigate them. The following year, Lewis reached out to the Global Priorities Institute to see if they'd be interested in employing him. He received a job offer and worked there from Jan-June 2019 learning more about longtermism.
After graduating summa cum laude (GPA 3.98), in the fall of 2019 Lewis started a PhD in economics at Stanford, where he planned to focus on the economics of AI, or other questions raised in the GPI research agenda. However, our headhunting team reached out to him about a role at DeepMind, and three months later he took an extended leave of absence from the PhD to work there as a Research Associate - where he examines a diverse mix of strategy related questions.
Lewis thinks that without 80,000 Hours he may not have gone to GPI or become as interested in AI safety or global priorities research, because at the time Yale EA events didn’t focus much on longtermism - and he wasn't reading much EA content apart from 80,000 Hours. He may have also not come across the DeepMind job he now has.
(We also learned that Zac Kenton - who we featured in our ‘Story of a plan change’ profile from bout 1 2019 - has recently left his position as a postdoc in machine learning at Oxford University to also join DeepMind as a researcher, focused on technical safety rather than strategy.)
Key lead metric targets from bout 2 (14th April - 30th June)
We made all our lead metric targets:
Key progress over bout 2 (April 14th to June 30th)
Overall, it seems like we had an unusually productive bout, meeting and often significantly exceeding our targets, which I’m happy about considering the situation with COVID and being fully remote.
More info by team:
I focused on investigating key uncertainties in our strategy, advice and EA in general. I feel like I learned a lot, and have written 15 internal docs, published 4 articles and drafted 9 more. I will aim to publish more of these over bout 3, and also incorporate them into an update of the key ideas page. I’ve listed some of the things I think I’ve learned below - feel free to contact me if there’s one you’re especially curious about.
Assessed Luisa Rodriguez and offered her a role. Previously Luisa has researched nuclear conflict with Rethink Priorities and is currently working with Will MacAskill on a book about longtermism. She plans to join us after the first draft is finished in mid-2021.
Released 13 pieces of content - including this overview of problem areas which aren’t currently our top priority areas (now the 4th most upvoted article on the EA forum) and this roundup of our 15 part series of anonymous advice.
So far over 2020, we’ve published 44 items, compared to 33 over all of 2019 (though with one extra staff member). See a list of all content we’ve released over 2020.
~24,000 unique clicks through to job vacancy pages via the job board (above target).
Released a job board user guide/FAQ to make it more understandable to new users,updated the job filters and the start page design.
We now have ~400 active listings (while ~200 was more typical in 2019).
Continued to experiment with ‘person-first’ as opposed to ‘role-first’ headhunting, and as part of this spoke to 80 people (target of 50) and sent 111 copies of The Precipice. We’ll give an update on which seems most promising in our annual review.
Person-first work starts with trying to identify promising people and seeing where they fit in; while role-first work starts with specific opportunities and then finds people to fill them.
Also continued to centralise our data, and outsourced much of the work.
Habiba Islam joined the team in May and is now advising independently (though still discussing each case with Michelle).
Advised 54 people (relative to a target of 44).
Made progress on strategy, investigating alternative versions of advising, the possible negative impact of rejections on the community, and the main critiques of advising by our stakeholders.
Planned 2020 impact evaluation, and think we worked out several ways to make it more accurate and take less time.
Oversaw ongoing office fitout. Unfortunately this has been highly delayed due to COVID, problems with contractors and a delay getting planning permission. We now expect it to be finished in Nov.
Some things I think I learned in the last two months of research
Here’s just a list of points - let me know if there’s something you’d especially like more detail on.
Estimated how the EA portfolio is allocated between causes, and now vs. later. Concluded current allocation is reasonable, except (i) there’s a gap in the portfolio around exploring broad longtermism interventions (e.g. reducing great power conflict) and (ii) global health seems over-allocated towards. Intend to promote broad longtermist options at 80k more going forward.
I reviewed our strategy towards near termist issues, climate change, and patient longtermism, and concluded that our current strategy is reasonable, though plan to tilt more towards patient longtermism compared to the past.
Considered how our advice might be wrong, and concluded that two of the biggest potential downside risks are getting the specialist vs. transferable career capital tradeoff wrong; and encouraging people out of second tier areas where they might be in the tail of success.
Categorised longtermism into 4 forms and made notes on the implication of each view for career priorities. Distinguished between broad vs. patient longtermism. Realised broad vs. patient longtermism are different; and realised that patient longtermism doesn’t obviously imply building transferable career capital (as I had assumed before).
Reviewed evidence on EA drop out rates, and made a new estimate using new data, and concluded that it’s lower than previous estimates for core community members. Investing a few years in building career capital in order to change your long term career trajectory (e.g. going to grad school) is more likely to pay off if drop out rates are low.
Realised the arguments for ‘earning to save’ being a significant part of the community in the future are better than I thought.
I learned a lot about now vs. later & predictors of personal fit / career success, though am still digesting what this means for our advice.
Did a quick analysis of how top EAs have gained career capital in the past; and of the best entry routes into our priority paths.
Drafted a new premise/conclusion argument for EA and intro to EA. This and the next bullet are outside of 80k’s central focus, but they seem really foundational to EA and no-one else is writing them.
Reviewed how much interventions differ in effectiveness, estimating that the boost from intervention selection compared to giving randomly is about 3x (less than I’d previously thought), making it much less important than cause selection.
Wrote a list of ways we might promote independent thinking (& how to balance with good judgement). I expect we’ll implement some of these ideas over the coming months.
Made a big list of ideas for changes to key ideas, which will aim to write next bout.
Plans for bout 3 2020 (6 weeks)
Research: release 8 pieces of content (and get 10,200 hours of engagement with content released during the bout) – focus these on either continuing to grow the podcast or filling gaps in the key ideas series.
Job board: on-going listing; maintain 1500 clicks/week moving average for at least 4/6 weeks of the bout; improving automation of listings (think can halve time per listing).
Headhunting: piloting a potential move into a more scalable CRM and doing 33 calls.
Advising: wrapping up strategy and Habiba on delivery, aiming to do 40 calls.
Impact evaluation: implementing the annual user survey & preparing for the annual review.
Ops: overseeing office refurbishment, hiring an office manager.
Till the next update,