Hide table of contents

If someone wants to learn about AI governance as part of their career exploration, they might have these problems:

  • It can be hard to know what resources to start with;
  • It can be hard to get a broad overview of the field (rather than understanding just eg. the law, or economics approaches);
  • Even if someone works out what they want to read, and it will give a broad overview, it can be hard for them to actually get through all the reading;
  • It can be hard to form your own views and resolve uncertainties by yourself.

EA Oxford has just finished running a reading group on AI governance, and it seems to have been successful at overcoming these problems. This was with low effort from the organisers, and it seems like it could be scalable. This post is for group organisers considering running an AI governance reading group.

Reading List

We used this syllabus put together by Markus Anderljung at the Centre for the Governance of AI, which is split up into nine topics, with 2-3 readings per topic. A lot of the value of the group came from having a syllabus that had a good balance of depth and breadth, and that came from someone with a really good understanding of the field. The list is useful to people interested in both research and implementation, and probably more useful to people interested in research.

Setting up the Group

We advertised the reading group to group members who had been through our In-Depth Fellowship (we run an Intro Fellowship similar to the Arete Fellowship, and then an In-Depth Fellowship). Having a small number of keen, core people made running the group very easy, and I was always happy with the quality of discussion.

We didn't advertise particularly widely because I was new to the content and running the group with core members seemed easier and less risky.

If we run the group again, we're interested in using it as an outreach tool to students in relevant subjects who haven't heard much about EA (and might be more attracted to 'making sure AI is governed well in the long-run' than 'doing the most good we can'). It seems like the tone of advertising and facilitation would be more important in this case as this would be some people's first large exposure to EA, and they could form a negative impression of the movement. It also seems more important for the organiser to have a good understanding of the content in this case, so that EAs don't seem incompetent or overconfident. I'd be excited for other groups to carefully try this.

Running the Group

We went through one topic per week and aimed to do all the non-further reading each week. This was around a two hour commitment. Meeting weekly seemed like a way to have time to do the reading, while also keeping momentum and a good social atmosphere.

While doing the reading each week, we wrote down these things in a shared document:

  • Our most interesting, surprising, or important updates from the reading;
  • Confusions, questions or uncertainties we had about the reading or the topic;
  • Anything we’d wanted to discuss eg. implications of the reading, extensions to the reading, thoughts we’d had over the week.

During the one hour long calls, we summarised each of the pieces from the reading list, then discussed what we'd written down.

The discussion was participant-led, and in general people seemed to be forming their own inside views and were happy to critique the reading.

Results

There've been five attendees including me who've come nearly every week, and around ten different attendees in total. The timing of the group clashed with exam season, and we also advertised a group that met monthly instead of weekly, so it seems like there could have been lots more attendees if we'd pushed for them. Some specific results of the group are that:

  • One person who will be doing the FHI Summer Research Fellowship came every week. They were initially planning to research a question on AI forecasting for the Summer Research Fellowship. They now think they'll do something in that space, but probably work on a different question as a result of the reading group.
  • One person who's a CS DPhil affiliate at FHI came almost every week. They found it useful to get a better understanding of the governance side of things, which they may transition into from work on technical AI safety.
  • The organiser (me), has a better understanding of AI governance problems - which I expect to be useful for explaining cause prioritisation, and giving career guidance.

One of the participants deprioritized AI Governance as a career path as a result of the group, thinking that their personal fit wasn't exceptional (which I see as a success of the group).

After the Group

A subset of the group will keep meeting weekly, with one person each week presenting on a relevant paper. This seems like a good way to dig deeper into the field. I'd also encourage group organisers to talk 1-1 to people who finish the group about career plans and next steps.

Further Reading

80,000 Hours - US AI Policy

AI Governance Career Paths for Europeans (implementation focussed)

Personal Thoughts on Careers in AI Policy and Strategy (research focussed)

80,000 Hours - Guide to Working in AI Policy and Strategy

Comments2


Sorted by Click to highlight new comments since:

Do you have a template of the shared document that you used? Or was it a quite unstructured blank document?

It's pretty blank - something like this

Curated and popular this week
 ·  · 5m read
 · 
This work has come out of my Undergraduate dissertation. I haven't shared or discussed these results much before putting this up.  Message me if you'd like the code :) Edit: 16th April. After helpful comments, especially from Geoffrey, I now believe this method only identifies shifts in the happiness scale (not stretches). Have edited to make this clearer. TLDR * Life satisfaction (LS) appears flat over time, despite massive economic growth — the “Easterlin Paradox.” * Some argue that happiness is rising, but we’re reporting it more conservatively — a phenomenon called rescaling. * I test rescaling using long-run German panel data, looking at whether the association between reported happiness and three “get-me-out-of-here” actions (divorce, job resignation, and hospitalisation) changes over time. * If people are getting happier (and rescaling is occuring) the probability of these actions should become less linked to reported LS — but they don’t. * I find little evidence of rescaling. We should probably take self-reported happiness scores at face value. 1. Background: The Happiness Paradox Humans today live longer, richer, and healthier lives in history — yet we seem no seem for it. Self-reported life satisfaction (LS), usually measured on a 0–10 scale, has remained remarkably flatover the last few decades, even in countries like Germany, the UK, China, and India that have experienced huge GDP growth. As Michael Plant has written, the empirical evidence for this is fairly strong. This is the Easterlin Paradox. It is a paradox, because at a point in time, income is strongly linked to happiness, as I've written on the forum before. This should feel uncomfortable for anyone who believes that economic progress should make lives better — including (me) and others in the EA/Progress Studies worlds. Assuming agree on the empirical facts (i.e., self-reported happiness isn't increasing), there are a few potential explanations: * Hedonic adaptation: as life gets
 ·  · 38m read
 · 
In recent months, the CEOs of leading AI companies have grown increasingly confident about rapid progress: * OpenAI's Sam Altman: Shifted from saying in November "the rate of progress continues" to declaring in January "we are now confident we know how to build AGI" * Anthropic's Dario Amodei: Stated in January "I'm more confident than I've ever been that we're close to powerful capabilities... in the next 2-3 years" * Google DeepMind's Demis Hassabis: Changed from "as soon as 10 years" in autumn to "probably three to five years away" by January. What explains the shift? Is it just hype? Or could we really have Artificial General Intelligence (AGI)[1] by 2028? In this article, I look at what's driven recent progress, estimate how far those drivers can continue, and explain why they're likely to continue for at least four more years. In particular, while in 2024 progress in LLM chatbots seemed to slow, a new approach started to work: teaching the models to reason using reinforcement learning. In just a year, this let them surpass human PhDs at answering difficult scientific reasoning questions, and achieve expert-level performance on one-hour coding tasks. We don't know how capable AGI will become, but extrapolating the recent rate of progress suggests that, by 2028, we could reach AI models with beyond-human reasoning abilities, expert-level knowledge in every domain, and that can autonomously complete multi-week projects, and progress would likely continue from there.  On this set of software engineering & computer use tasks, in 2020 AI was only able to do tasks that would typically take a human expert a couple of seconds. By 2024, that had risen to almost an hour. If the trend continues, by 2028 it'll reach several weeks.  No longer mere chatbots, these 'agent' models might soon satisfy many people's definitions of AGI — roughly, AI systems that match human performance at most knowledge work (see definition in footnote). This means that, while the compa
 ·  · 4m read
 · 
SUMMARY:  ALLFED is launching an emergency appeal on the EA Forum due to a serious funding shortfall. Without new support, ALLFED will be forced to cut half our budget in the coming months, drastically reducing our capacity to help build global food system resilience for catastrophic scenarios like nuclear winter, a severe pandemic, or infrastructure breakdown. ALLFED is seeking $800,000 over the course of 2025 to sustain its team, continue policy-relevant research, and move forward with pilot projects that could save lives in a catastrophe. As funding priorities shift toward AI safety, we believe resilient food solutions remain a highly cost-effective way to protect the future. If you’re able to support or share this appeal, please visit allfed.info/donate. Donate to ALLFED FULL ARTICLE: I (David Denkenberger) am writing alongside two of my team-mates, as ALLFED’s co-founder, to ask for your support. This is the first time in Alliance to Feed the Earth in Disaster’s (ALLFED’s) 8 year existence that we have reached out on the EA Forum with a direct funding appeal outside of Marginal Funding Week/our annual updates. I am doing so because ALLFED’s funding situation is serious, and because so much of ALLFED’s progress to date has been made possible through the support, feedback, and collaboration of the EA community.  Read our funding appeal At ALLFED, we are deeply grateful to all our supporters, including the Survival and Flourishing Fund, which has provided the majority of our funding for years. At the end of 2024, we learned we would be receiving far less support than expected due to a shift in SFF’s strategic priorities toward AI safety. Without additional funding, ALLFED will need to shrink. I believe the marginal cost effectiveness for improving the future and saving lives of resilience is competitive with AI Safety, even if timelines are short, because of potential AI-induced catastrophes. That is why we are asking people to donate to this emergency appeal