80k supporter update - bout 3 and 4 review
Benjamin Todd
Hi everyone,
Here’s our update since June, summarising our progress over bout 3 and 4 (28th May - 27th Aug), along with our plans for bout 5. See our previous update.
Progress in 2 sentences
We roughly doubled the number of jobs on the job board, which appears to be helping growth. We also onboarded Jenna Peters as our second advisor, advised 56 people, made 30 headhunting introductions to organisations, released 7 pieces of content, did a mid-year review of strategy and attended the Effective Altruism Leaders Forum.
Summary metrics
Growth metric:
Impact-adjusted significant plan changes (IASPC) - see more explanation.
Annual IASPC targets
All time IASPC growth chart
Note we think the relative weightings of the IASPC scale are not correct and should be more stretched out. There are also several other problems with our plan change metric. We plan to update our system this year, and will post an update about this here when it’s done.
-----------------------Full update -----------------------
Story of a plan change
Carina studied a PhD at Oxford University in the philosophy of quantum mechanics, and initially planned to become an academic in the area. She was working in a philosophy study room when she came across the 80,000 Hours book, which someone had left unattended on a desk. She read the book, then our our online materials, and through this learned about effective altruism and the importance of AI governance.
Carina applied for one-on-one advice and spoke to Brenton (an advisor at 80,000 Hours), who suggested she apply to the research scholars program. She was accepted, and is now a Senior Research Scholar working on the Ethics and Governance of Artificial Intelligence.
Some updates on the job board
This year we increased investment in our job board, bringing Maria up to 80% time on the project, with support from the tech team.
One aim has been to expand the range of roles on the board. She recently worked with Niel Bowerman (our AI policy specialist) to generate a list of promising organisations in the area, and criteria for deciding what jobs to list. As a result of their work, we're currently advertising around 200 roles related to AI strategy and governance, including some entry-level roles to make it easier to enter the area. We hope this is now at the level where a reasonably well-qualified person interested in the area will generally find several new promising roles.
Initial results suggest that listing more jobs leads to more people clicking through to the external listings, especially when we release job board newsletters (which are the spikes below). We take clicking through to the ads to be an indication that people find the jobs interesting, which hopefully predicts placements. We will do a more detailed analysis of placements over the coming months as part of our impact evaluation. If this is positive, we think we could significantly expand the number of jobs we list in other areas, and further grow the impact of the board.
(Note that the spikes in the chart correspond to when we send our newsletter.)
We continue to face challenges with accurately conveying our views on the promise of each job, which have become more pressing as we’ve listed a wider range of them. One example is that jobs we list mainly for gaining career capital can seem odd without further explanation, and might get confused for positions we think are high-impact (e.g. we list public policy positions at Amazon). We do flag these jobs as for skill-building, but it’s easy to miss, so we’ll work on it more over the coming months.
Key progress over bouts 3 and 4
In this period, we did some work on strategy:
Most importantly, we attended the Effective Altruism Leaders Forum, where we learned about the problems facing EA people think are most pressing, and thought about how 80k might help to solve them.
One message was that doing more headhunting as a way to help solve the ‘talent integration’ bottleneck was popular among attendees. We’ve also been thinking about how to highlight more options for gaining career capital, to make it easier to enter top areas.
The topic that was most discussed could perhaps be categorised as culture. Some worries about the culture that were expressed included it being too unidimensional, inward-facing, and overconfident in its frameworks. We intend to keep working on our online content to improve on these points e.g. expressing more caution, finding ways to feature a wider variety of paths, and hosting podcast guests who exemplify good culture.
We also did a mid-year ‘strategy week’ where the team reflected on bigger picture issues facing their programmes that we don’t usually get a chance to think about week-to-week. Some of the topics included: how we should update our views on our impact based on an external analysis; how much to downgrade our rating of AI safety based on Ben Garfinkel’s research, and how we can better communicate confidence in our online content - potentially allowing us to publish more often.
We also made steady progress on delivering our core programmes:
Online content: we released 8 pieces of content and I was able to make progress on articles for our key ideas series during a two week writing retreat. We released:
Pushmeet Kohli on DeepMind’s plan to make AI systems robust & reliable
Philip Tetlock on predicting the future better and cross-promotion of the Open Phil / Clearer Thinking calibration tool
Paul Christiano on messaging the future, increasing compute, & how CO2 impacts your brain.
Helen Toner on emerging technology, national security, and China.
Rob also appeared as a guest on the Neoliberal Podcast and Love Your Work.
Advising: we onboarded our new advisor, Jenna Peters, and advised 56 people.
Headhunting: Niel completed his first full search, and we learned an introduction he made in an earlier bout resulted in a new hire. We also completed several other searches and made more than 30 introductions to key organisations working in our priority paths (who have historically been hired at a rate of ~1 in 11, though this varies widely by organisation and role).
Web and tech: we advertised over 300 vacancies on the job board, added over 100 organisations to screen for vacancies, and worked on other features.. We released an internal tool for gathering all our data on each user, completed an organisational security audit, and published a round of edits to the front page.
We made our lead metric targets for the bouts, except we missed the content release target in bout 4. That was in significant part due to some unanticipated delays finishing content (e.g. realising we needed to rewrite something, waiting on comments). When making the bout 4 goal, we also didn’t properly take into account that we were working on more written articles compared to podcasts than normal, and that these typically take longer.
We’re now in bout 5, during which we’re gathering materials for our annual review and fundraising round. In the remaining time, we’ll keep working on delivering our programmes. So far we’re just about on track with our metrics for bout 5.
Till the next update,
Ben
