I am a rising junior at the University of Chicago (co-president of UChicago EA and founder of Rationality Group). I am mostly interested in philosophy (particularly metaethics, formal epistemology, and decision theory), economics, and entrepreneurship.
I also have a Substack where I post about philosophy (ethics, epistemology, EA, and other stuff). Find it here: https://substack.com/@irrationalitycommunity?utm_source=user-menu.
Reach out to me via email @ dnbirnbaum@uchicago.edu
This might feel obvious, but I think it’s under-appreciated how much disagreement on AI progress just comes down to priors (in a pretty specific way) rather than object-level reasoning.
I was recently arguing the case for shorter timelines to a friend who leans longer. We kept disagreeing on a surprising number of object-level claims, which was weird because we usually agree more on the kinda stuff we were arguing about.
Then I basically realized what I think was going on: she had a pretty strong prior against what I was saying, and that prior is abstract enough that there’s no clear mechanism by which I can push against it. So whenever I made a good object-level case, she’d just take the other side — not necessarily because her reasons were better all else equal, but because the prior was doing the work underneath without either of us really knowing it.
There’s something clearly rational here that’s kinda unintuitive to get a grip on. If you have a strong prior, and someone makes a persuasive argument against it, but you can’t identify the specific mechanism by which their argument defeats it, you should probably update that the arguments against their case are better than they appear, even if you can’t articulate them yet. From the outside, this totally just looks like motivated reasoning (and often is), but I think it can be pretty importantly different.
The reason this is so hard to disentangle is that (unless your belief web is extremely clear to you, which seems practically impossible) it’s just enormously complicated. Your prior on timelines isn’t an isolate thing — it’s load-bearing for a bunch of downstream beliefs all at once. So the resistance isn’t obviously irrational, it’s more like… the system protecting its own coherence.
I think this means that people should try their best to disentangle whether some object level argument they’re having comes from real object level beliefs or pretty abstract priors (in which case, it seems less worthwhile to press on them).
The Forum should normalize public red-teaming for people considering new jobs, roles, or project ideas.
If someone is seriously thinking about a position, they should feel comfortable posting the key info — org, scope, uncertainties, concerns, arguments for — and explicitly inviting others to stress-test the decision. Some of the best red-teaming I’ve gotten hasn’t come from my closest collaborators (whose takes I can often predict), but from semi-random thoughtful EAs who notice failure modes I wouldn’t have caught alone (or people think pretty differently so can instantly spot things that would have taken me longer to figure out).
Right now, a lot of this only happens at EAGs or in private docs, which feels like an information bottleneck. If many thoughtful EAs are already reading the Forum, why not use it as a default venue for structured red-teaming?
Public red-teaming could:
reduce unilateralist mistakes,
prevent coordination failures (I’ve almost spent serious time on things multiple people were already doing — reinventing the wheel is common and costly),
Obviously there are tradeoffs — confidentiality, social risk, signaling concerns — but I’d be excited to see norms shift toward “post early, get red-teamed, iterate publicly,” rather than waiting for a handful of coffee chats.
Here’s a random org/project idea: hire full-time, thoughtful EA/AIS red teamers whose job is to seriously critique parts of the ecosystem — whether that’s the importance of certain interventions, movement culture, or philosophical assumptions. Think engaging with critics or adjacent thinkers (e.g., David Thorstad, Titotal, Tyler Cowen) and translating strong outside critiques into actionable internal feedback.
The key design feature would be incentives: instead of paying for generic criticism, red teamers receive rolling “finder’s fees” for critiques that are judged to be high-quality, good-faith, and decision-relevant (e.g., identifying strategic blind spots, diagnosing vibe shifts that can be corrected, or clarifying philosophical cruxes that affect priorities).
Part of why I think this is important is because I generally think have the intuition that the marginal thoughtful contrarian is often more valuable than the marginal agreer, yet most movement funding and prestige flows toward builders rather than structured internal critics. If that’s true, a standing red-team org — or at least a permanent prize mechanism — could be unusually cost-effective.
There have been episodic versions of this (e.g., red-teaming contests, some longtermist critiquing stuff), but I’m not sure why this should come in waves rather than exist as ongoing infrastructure (org or just some prize pool that’s always open for sufficiently good criticisms).
Seems right, though plausibly there are some EA/EA-adjacent students at some of these schools.
Why don’t EA chapters exist at very prestigious high schools (e.g., Stuyvesant, Exeter, etc.)?
It seems like a relatively low-cost intervention (especially compared to something like Atlas), and these schools produce unusually strong outcomes. There’s also probably less competition than at universities for building genuinely high-quality intellectual clubs (this could totally be wrong).
Thanks for the comment and good points.
What I meant is that they can be MORE politically charged/mainstream/subject to motivated reasoning. I definitely agree that current incentives around AI don’t perfectly track good moral reasoning.
Yep, I agree (though I’m not sure if I agree that the incentive is clearly in the negation; one could argue that a company may want to say that they are worried about sentience to increase hype the same way some argue that talking about the risks of AI increases hype). I just think there will be more when the issue is in the minds of the public.
I think there are some mainstream things about digital minds (Black Mirror comes to mind), but I don’t think it’s a thing that people yet take seriously in the real world.
Yea, unclear if these self-reports will be reliable, but I agree that this could be true (and I briefly mention something like it: “Broadly, AW has high tractability, enormous current scale, and stronger evidence of sentience—at least for now, since future experiments or engineering relevant to digital minds could change this.”
I think having new places to make more hits based bets (within orgs or creating new orgs) is going to be really important.
My sense is that we have somewhat of a grantmaker shortage now (though correct me if I’m wrong), and this problem will become worse.
I guess that I am just not very confident that this is a good proxy.
The allocation is probably not what it should be, but I’m not yet convinced that animal stuff is being underfunded.
This seems to assume that EA funds ought to be distributed “democratically” by people who identify as EAs or EA leaders. I don’t buy that.
If the goal is good resource allocation, we want funding decisions to track competence about cause prioritization, not representativeness. Randomly sampled EAs—many of whom haven’t spent much time thinking seriously about cause prioritization—don’t seem like the right decision-makers.
And it’s also not obvious that “EA leaders,” as such, are the right allocators either. The relevant property isn’t status or identity, but judgment, track record, and depth of thinking about cause prioritization.
I think one can reasonably ask this question of consciousness/welfare more broadly: how does one have access to their consciousness/welfare?
One idea is that many philosophers think one, by definition, has immediate epistemic access to their conscious experiences (though whether those show up in reports is a different question, which I try to address in the piece). I think there are some phenomenological reasons to think this.
Another idea is that we have at least one instance where one supposedly has access to their conscious experiences (humans), and it seems like this shows up in behavior in various ways. While I agree with you that our uncertainty grows as you get farther from humans (i.e. to digital minds), I still think you’re going to get some weight from there.
Finally, I think that, if one takes your point too far (there is no reason to trust that one has epistemic access to their conscious states), then we can’t be sure that we are conscious, which I think can be seen as a reductio (at least, to the boldest of these claims).
Though let me know if something I said doesn’t make sense/if I’m misinterpreting you.
I would love to hear any updated takes on this post from Ryan.
I agree this is a super hard problem, but I do think there are somewhat clear steps to be made towards progress (i.e. making self reports more reliable). I am biased, but I did write this piece on a topic that touches on this problem a bit that I think is worth checking out.
Vasco, you are the king of cross-posting good content onto the EA forum.
Yea, I agree with this a lot.
One thing to warn, though: one can be inclined to say something like “I’m an actor, so I don’t have to do any of the accuracy-based forecasting, etc,” which I think is definitely wrong.
One should choose which direction they act based on some somewhat accuracy/EV + vibes. Otherwise, there are infinite ways to act and most of them, without any foresight, fail even with the type of agency you’re describing.
In addition, once you are in the project you are acting in, you (ofc) shouldn’t constantly be doing the accuracy/EV thing. Sometimes, though, it’s probably worth taking a step back and doing it to consider opportunity cost (on some marked date every x amount of time to avoid decision/acting paralysis).
Dwarkesh (of the famed podcast) recently posted a call for new guest scouts. Given how influential his podcast is likely to be in shaping discourse around transformative AI (among other important things), this seems worth flagging and applying for (at least, for students or early career researchers in bio, AI, history, econ, math, physics, AI that have a few extra hours a week).
The role is remote, pays ~$100/hour, and expects ~5–10 hours/week. He’s looking for people who are deeply plugged into a field (e.g. grad students, postdocs, or practitioners) with high taste. Beyond scouting guests, the role also involves helping assemble curricula so he can rapidly get up to speed before interviews.
More details are in the blog post; link to apply (due Jan 23 at 11:59pm PST).
Hmm this is interesting.
Speculatively, I think there could actually just be convergence here, though, once you account for moral uncertainty and just very plausible situations where doing bad by everyone’s lights are as bad as, say, utilitarian nightmares but just easier to get others on board for (ie extreme power).
Some posts are meant for literary beauty and some are meant for ideas. Writing ideas well takes effort, and writing ideas with AI takes basically zero effort. If your post is mainly about ideas, and the fact that it is written by an AI doesn’t make it annoying to read, I really just don’t see why you wouldn’t write basically entirely with AI.