Announcing: 'EA and AI' Debate Week

TLDR:

• Between 30 Nov – 6 Dec we’ll debate the question: “What % of EA resources should go to making AI go well?”. This question is disambiguated below.
• 80,000 Hours advisors will host an Ask Me Anything thread during the week for anyone considering whether to make a career shift into making AI go well.
• You are invited to write a post for the week - or otherwise engage in debate during the event. 

Thanks to @brentonmayer for nudging me to host this debate week. Clearly, the question of just how much EA as a movement should be investing in making AI go well is particularly pertinent at the moment, as more funding is (possibly) on its way, AI safety work is becoming more and more mainstream, and AI capabilities continue to advance. 

I want debate weeks like this to be a chance for the community of EAs who are usually heads-down and focused on their work to look up, and consider whether they are using their resources in line with their values. Personally, I'm uncertain on this question. I see the case for urgency – but urgency can mask low tractability. And I'm always prima facie sceptical when EA – the impact avant-garde – sticks around once a cause becomes more mainstream. 

Brenton and some staff from Rethink Priorities^[1] have committed to writing something for the event. A few words from Brenton to kick us off:

Brenton (80,000 Hours):

The question of effective altruism is “How can I do the most good, with the resources available to me?” In 2026, I think the answer for most people is “something to help make AI go well”.

I’m unsure what fraction of EA resources is right. But I expect the share going to making AI go well should be higher than what we thought in 2024.

Getting the EA ecosystem better allocated will mean lots of people checking in on whether they’re in the right place to help others as best as they can. I’d love for people to use the Forum debate week as a prompt to do so: by reading Rethink’s work, making use of the 80,000 Hours advisor AMA, and reading other perspectives which folks contribute during the week.

In my post I’ll go through what I think are the most important updates for people’s decisions: that the case for AI risk has become much less speculative, and that people have a better chance of finding a role in making AI go well that suits them than they used to.

Clarifications and disambiguations [don’t skip]

“During the week, we will discuss what percentage of EA resources [1] should be spent [2] on making AI go well [3]”.

1. ‘EA resources’: This is, of course, a fuzzy category. We do not own our donors money, and foundations which are mostly EA aligned may also decide to give to far less effective interventions. Where you draw the boundaries is up to you, but my proposed definition is ‘resources controlled by impartial altruists’. This includes labour. It’s reasonable to focus your vote or post on only one of labour or funding - your vote on both is likely to be pretty similar, and it’s best to vote based on what you know best.
2. ‘Should be spent’: Let’s bound this to 2027 spending. Take the total EA resources that will be spent in 2027 as fixed: 100% means all of it goes to making AI go well. The question of how fast to spend down reserves of funding is a separate one.
3. ‘Making AI go well’: I’ve gone for this purposefully broad statement (rather than ‘AI safety’) to encompass a range of work that assumes societal transformation from AI, inclusive of but not limited to risks of extinction. For this event, digital minds and better futures work count as 'making AI go well'. 

As always, your vote won’t be perfectly unambiguous. When you vote, you’ll be prompted to leave a linked comment — that’s your chance to explain how you understand the debate question, and what you take your vote to mean.

What to expect during the event

During debate weeks, the centrepiece is a banner at the top of the EA Forum homepage, where logged-in users can vote along a spectrum. It looks a bit like this:

When you vote, you'll be prompted to leave a comment in a discussion thread, which then encourages conversation about the issue. The best outputs of debate weeks, however, are the posts. You could write one.

Debate week posts will be highlighted during the week, in a separate frontpage tab on the Forum. If you feel strongly about this debate question, this is a great opportunity to test your viewpoints, and have some influence on how the EA community thinks.

What counts as a debate week post? Any post which contains information or an argument which could reasonably be expected to update an impartial altruist on the central debate question. In other words, your post doesn’t have to be a debate-style argument to be a great debate week post. Great posts could be:

• A post sharing a tool that tracks EA-aligned grants and careers.
• An argument that the recipients of EA focus — the global poor, animals, future people — are best served by working on helping the transition to AGI go well.
• Your own model of how much funding is likely to be deployed next year and where.
• An argument that, due to cluelessness, we should not invest further in making AI go well and should instead focus on near-term more certain routes to impact.

Pre-reading, for developing your takes

• Background information on the distribution of funding, now and in the future:
  • Metaculus predictions on how much the AI windfall will be, and where it will go.
  • A (caveated) database of EA-aligned grants.
  • The most recent gauge of EA cause prioritisation thinking: EA Survey 2024: Cause prioritization.
• Giving and AGI: Why “just AI now, save rest for later” is a risky plan, Hayley Clatterbuck (Rethink Priorities)
• Do short AI timelines make other cause areas useless?, Hayley Clatterbuck, arvomm
• Causes and Uncertainty: Rethinking Value in Expectation, Bob Fischer (Rethink Priorities)
• Effective altruism in the age of AGI, William MacAskill
• 80,000 Hours is shifting its strategic approach to focus more on AGI, 80,000 Hours, Niel Bowerman
• Animal Welfare is Just Part of AI Alignment Now, Aidan Kankyoku
• EA is underfunding animal advocacy according to our own preferences, ElliotTep
• Recommend additions to the reading list in the comments. 

Feedback? Questions? Leave your comments below.

1. ^{^}

   Rethink Priorities says that the "Rethink Priorities team will discuss the factors that should determine the answer to this type of allocation question and examine the reasoning behind their recommendation based on the latest version of the cross-cause model (updates about to be published). They may also comment on other takes during the week."