In EA, people use the word "counterfactual" in a non-standard way, but I've never heard this discussed or pointed out. E.g. Jeff writes,
Say I offer to make a counterfactual donation of $50 to the Against Malaria Foundation (AMF) if you do a thing; which of the following are ok for me to do if you don't?
But outside of the community, "counterfactual" means "didn't happen". I think the word "causal" is closest in meaning to how we use "counterfactual," though it doesn't work in this case.
(In this case, I think the standard English way of communicating the idea is "Say I offer to make a donation of $50 to AMF only if you do a thing...")
In Bruce Friedrich's new book, he writes, "Sometimes when I talk about cultivated meat someone will bring up the handful of states that have banned it. I'm mostly unconcerned. Cultivated meat companies won't be able to supply all 50 US states anytime soon anyway. Once there are multiple companies selling their products in...the majority of cities all across the country, the states that banned will-- I predict-- quietly repeal their laws" (p. 191).
It's hard to know how literally to interpret this apparently sanguine attitude, as the book is designed to generate enthusiasm for alternative proteins. But, still it seems raise an important question about the cost-effectiveness of repealing existing bans or preventing further ones. Initial thoughts:
Applications are now open for TARA Round 2 2026!
Apply by July 26th to join our 14-week APAC-based technical AI safety program (using the ARENA curriculum), designed to accelerate your path to meaningful technical alignment research.
TARA is built for you to learn around full-time work or study. You'll attend in-person sessions in your home city on Saturdays and do independent study throughout the week. You'll finish the program with a project to add to your portfolio, key technical AI safety skills, and connections across APAC.
In this Round we are targeting Melbourne, Sydney, Brisbane, Tokyo, Manila, Singapore, Taipei, Hong Kong, Bangalore, Delhi, Hyderabad and Auckland. Which cities run will depend on where applicant demand and strength are highest.
Our results so far:
→ 90% likelihood-to-recommend
→ 95% very or extremely satisfied with program management
→ 91% more likely to pursue a career in AI safety
→ 60% of surveyed 2025 graduates secured a fellowship or AI safety role within a year (9 of 15 tracked)
Learn more about TARA and apply through our website here.
Strong upvote - too many people see CEA as an authority source on everything EA, rather than a bunch of staff the EA community pays for to do the safe, repeatable stuff.
I'm excited to see this happen! Some welfare tech ideas seem really promising, and I think this is clearly a promising project — it's great that it's happening.
My biggest areas of uncertainty about welfare tech (which overlap with some you flagged and some of which apply to only some technologies):
Thanks so much for doing this! Seems like a great thing to try, and exciting to see the technology that will come out of it!
I was never accepted into the Charity Entrepreneurship programme, but I was able to land a high-impact role at one (now leading it!), by getting in touch with other founders. For me, this is no less rewarding than being the real founder, and enough people treat me as if I went through the programme that it’s usually a surprise to people that I didn’t.
So I would urge some people to reconsider if founding is exactly what you want, you may be able to derive almost all of the benefits via a slightly different path, and CE’s charities would very much welcome talented founder-type generalists in high-level leadership & ownership roles!
Wait back to 50% because it says “end factory farming” rather than “ substantially reduce factory farming”
Here’s Holden Karnofsky:
I tend to think it’s worse than 51/49. I tend to think we’re always going to be prone to overestimate how robustly good our actions are. And the more we learn about all the galaxy-brained considerations that one should have had in one’s head, the more it’s going to be like 50+ε%. I think AI safety is a great cause to work in. I’m excited t...
Future life may turn out to be net negative (e.g. s-risks might become real), and work that preserves the chances of humanity spreading may essentially be enabling that.
In almost the opposite direction, AI might just be very useful and beneficial and a bunch of the risks people in AI safety worry about might not end up becoming real, or might be quite easy to overcome. And in the meantime, AI safety work may have:
Context: I wrote this post mainly to help me clarify my own thoughts on the topic. I don’t think the concerns I raise here are particularly novel (they should be discussed more though).
Thanks for this! Firstly these seem like two separate Qs (“which actors am I most worried about?” vs the specific risks posed by defender-first access programmes).
My current guess (as a relatively uninformed outsider) is that the most worrying near-term category actors with some combination of pre-existing domain knowledge, operational access, and (obviously) intent. So in bio I guess this would mean: state/state-adjacent programmes, small highly motivated teams with wet-lab access, and maybe compromised actors inside institutions that give them access. I definitely don't currently think frontier models are good enough that a literal random person could cause catastrophe if they wanted to (though this could change).
Re defender-first access specifically: yeah makes sense that this could add risk. E.g. if there's stuff like insider misuse, account compromise/coercion, leakage of outputs/workflows, etc.
I don't really have more concrete thoughts beyond this. It would be interesting to see some analysis (if there isn't already one internally) on how robust OAI's Rosalind Biodefense and Anthropic's Project Glasswing are to risks from risky actors getting into the programme.
(Somewhat relatedly I remember reading somewhere that Mythos reportedly got accessed by some unauthorised users via a third-party private Discord group or something).
Note: This post was crossposted from Planned Obsolescence by the Forum team, with the author's permission. The author may not see or respond to comments on this post.
But we get the job done
...
This is a follow-up to my (now fairly outdated) August 2025 post.
This post contains all the AI safety research and project ideas I've had in the past 10 months which I think could be high impact. I’m sharing them in case any are helpful or generative for others. I don’t plan to pursue most of these myself as there are just too many for m...
“used AI to help edit this post; all arguments were reviewed and modified by me."
Introduction
As participants in the Plateau State University School of Postgraduate Studies Effective Altruism Fellowship, we recently read William MacAskill’s Doing Good Better. Like many people encountering Effective Altruism for the first time, we initially thought the book would simply encourage charitable giving. Instead, we discovered something much broa...
Hi, thanks for this post and for your engagement :) If you haven't checked them out yet there are quite a lot of resources on similar topics, which may not answer all the questions you have but can be a good starting point (and can also help put you in touch with others who are dealing with similar things).
See for example this post from 2022, the comments and the resources it links to (there are probably more recent ones too):
https://forum.effectivealtruism.org/posts/Hyco4iMbL6phJwCH2/ea-career-guide-for-people-from-lmics
[Subtitle.] The case for Claude and ChatGPT having subjective experiences
This is a crosspost for Time to take AI consciousness seriously by Samuel Hammond, which was originally published on Second Best on...
I remain sceptical of computational functionalism (CF). It implies some sets of AND, OR, and NOT operations lead to consciousness even if run at the rate of one operation every billion years, which seems very implausible to me.
I used AI to fix transcription errors, rerrarange the ideas, and suggest tweaks to the title and some sentences.
Three of the most exciting projects to come out of EA in recent years are, in a vague sense, CEA spinouts:
Agree, I think people should "be the change they want to see" more often, especially in community / field building.
One caveat: Understand reputational harm risks and manage them (don't overestimate them either)
If your projects includes using the EA brand for outreach or other public comms, I'd recommend at least informing CEA or a local (full-time, paid) EA community builder, and asking if there's anything you should consider.
In some cases, there might be downside risks to other EA orgs or EA as a whole, some of which you might overlook, especially if you're new to EA community building or haven't yet gotten a lot of feedback. If that's the case, you can still do the project, maybe just not using the EA brand, or maybe test it small-scale first, with people who can give you honest feedback, before scaling up.
(this is just my personal opinion, informed by my time as EA Germany director until 2022, but not speaking for EA Germany or any other EA org)
What are the central virtues a committed utilitarian should cultivate? I would kill for a really great answer to this question
I think Stefan Schubert & Lucius Caviola's Virtues for Real-World Utilitarians is pretty good!
I agree! I was thinking of doing something like this and wrote an (unpublished) post which mentioned it, still strongly considering doing something related though perhaps more so a “flourishing futures”/“better futures” angle, which I think is related.
It looks like we had a pretty massive bot attack last night, with over 600 accounts logging in and messaging users about an iPad scam. Please don't click on any of their links. The scam looks like this:
We'll remove the accounts, and karma-gate messaging to 10 karma so that this cannot happen again. Sorry everyone - and thanks so much to everyone who reported this. I can't respond to all of you, but thank you.