Studying behaviour and interactions of boundedly rational agents, AI alignment and complex systems.
Research fellow at Future of Humanity Institute, Oxford. Other projects: European Summer Program on Rationality. Human-aligned AI Summer School. Epistea Lab.
I was considering hypothetical scenarios of the type “imagine this offer from MIRI arrived, would a lab accept” ; clearly MIRI is not making the offer because the labs don’t have good alignment plans and they are obviously high integrity enough to not be corrupted by relatively tiny incentives like $3b
I would guess there are ways to operationalise the hypothethicals, and try to have, for example, Dan Hendrycks guess what would xAI do, him being an advisor.
With your bets about timelines—I did 8:1 bet with Daniel Kokotajlo against AI 2027 being as accurate as his previous forecast, so not sure which side of the “confident about short timelines” do you expect I should take. I’m happy to bet on some operationalization of your overall thinking and posting about the topic of AGI being bad, e.g. something like “3 smartest available AIs in 2035 compare all what we wrote in 2026 on EAF, LW and Twitter about AI and judge who was more confused, overconfident and miscalibrated”.
The operationalisation you propose does not make any sense, Yudkowsky and Soares do not claim ChatGPT 5.2 will kill everyone or anything like that.
What about this:
MIRI approaches [a lab] with this offer: we have made some breakthrough in ability to verify if the way you are training AIs leads to misalignment in the way we are worried about. Unfortunately the way to verify requires a lot of computations (ie something like ARC), so it is expensive. We expect your whole training setup will pass this, but we will need $3B from you to run this; if our test will work, we will declare that your lab solved the technical part of AI alignment we were most worried about & some arguments which we expect to convince many people who listen to our views.
Or this: MIRI discusses stuff with xAI or Meta and convinces themselves their—secret—plan is by far the best chance humanity has, and everyone ML/AI smart and conscious should stop whatever they are doing and join them.
(Obviously these are also unrealistic / assume something like some lab coming with some plan which could even hypotehically work)
The financial basis for motivated reasoning is arguably even stronger in MIRI’s case than in Mechanize’s case. The kind of work MIRI is doing and the kind of experience Yudkowsky and Soares have isn’t really transferable to anything else.
It is somewhat difficult to react to this level of absolutely incredible nonsense politely, but I’ll try.
I disagree with both Yudkowsky and Soares about many things, but very obviously their direct experience with thinking and working with existing AIs would be worth > $1M pa if evaluated anonymously based on understanding SOTA AIs, and likely >$10s M pa if they worked on capabilities.
For the companies racing to AGI, Y&S endorsing some effort as good would likely have something between billions $ to tens of billions $ value.
Travel: mostly planned (conferences, some research retreats).
We expect closely coordinated team work on the LLM psychology direction, with a bit looser connections to the gradual disempowerment / macrostrategy work. Broadly ACS is small enough that anyone is welcome to participate in anything they are interested in, and generally everyone has idea what others work on.
My impression is EAGx Prague 22 managed to balance 1:1s with other content simply by not offering SwapCard 1:1s slots part of the time, having a lot of spaces for small group conversations, and suggesting to attendees they should aim for something like balanced diet. (Turning off SwapCard slots does not prevent people from scheduling 1:1, just adds a little friction; empirically it seems enough to prevent the mode where people just fill their time by 1:1s).
As far as I understand this will most likely not happen, because weight given to / goodharting on metrics like people reporting 1:1s is the most valuable use of time, metrics tracking “connections formed” and weird psychological effect of 1:1 fests. (People feel stimulated, connected, energized,… Part of the effect is superficial). Also the counterfactual value lost from lack of conversational energy at scales ~3 to 12ppl is not visible and likely not tracked in feedback (I think this has predictable effects on what types of collaborations do start and which do not, and the effect is on the margin bad.) The whole is downstream of problems like Don’t Over-Optimize Things / We can do better than argmax.
Btw I think you are too apologetic / self-deprecating (“inexperienced event organisers complaining about features of the conference”). I have decent experience running events and all what you wrote is spot on.
Thanks for explanation. My guess is this decision should not be delegated to LLMs but mostly to authors (possibly with some emphasis on correct classification in the UI).
I think the “the post concerns an ongoing conversation, scandal or discourse that would not be relevant to someone who doesn’t care about the EA community” should not be interpreted extensively, otherwise it can easily mean “any controversy or criticism”. I will repost it without the links to current discussions—these are non-central, similar points are raised repeatedly over the years and it is easy to find dozens of texts making them.
I wrote a post on “Charity” as a conflationary alliance term. You can read it on LessWrong, but I’m also happy to discuss it here.
If wondering why not post it here: Originally posted it here with a LW cross-post. It was immediately slapped with the “Community” tag, despite not being about community, but about different ways people try to do good, talk about charity & ensuing confusions. It is about the space of ideas, not about the actual people or orgs.
With posts like OP announcements about details of EA group funding or EAG admissions bar not being marked as community, I find it increasingly hard to believe the “Community” tag is driven by the stated principe marking “Posts about the EA community and projects that focus on the EA community” and not by other motives, like e.g. forum mods expressing the view “we want people to think less about this / this may be controversial / we prefer someone new to not read this”.
My impression this moves substantial debates about ideas to the side, which is a state I don’t want to cooperate on by just leaving it as it is → moved the post on LessWrong and replaced by this comment.
Seems plausible the impact of that single individual act is so negative that aggregate impact of EA is negative.
I think people should reflect seriously upon this possibility and not fall prey to wishful thinking (let’s hope speeding up the AI race and making it superpower powered is the best intervention! it’s better if everyone warning about this was wrong and Leopold is right!).
The broader story here is that EA prioritization methodology is really good for finding highly leveraged spots in the world, but there isn’t a good methodology for figuring out what to do in such places, and there also isn’t a robust pipeline for promoting virtues and virtuous actors to such places.
I don’t think so. I think in practice
I. - Some people don’t like the big R community very much.
AND
2a. - Some people don’t think improving the EA community small-r rationality/epistemics should be one of top ~3-5 EA priorities.
OR
2b. - Some people do agree this is important, but don’t clearly see the extent to which the EA community imported healthy epistemic vigilance and norms from Rationalist or Rationality-adjacent circles
=>
- As a consequence, they are at risk of distancing from small r rationality as a collateral damage / by neglect
Also I think many people in the EA community don’t think it’s important to try hard at being small-r rational at the level of aliefs. No matter what is the actual situation revealed by actual decisions, I would expect the EA community to at least pay lip service to epistemics and reason, so I don’t think stated preferences are strong evidence.
“Being against small-r rationality is like being against kindness or virtue; no one thinks of themselves as taking that stand.”
Yes I do agree almost no one thinks about themselves that way. I think it is maybe somewhat similar to “Being against effective charity”—I would be surprised if people though about themselves that way.
Reducing rationality to “understand most of Kahneman and Tversky’s work” and cognitive psychology would be extremely narrow and miss most of the topic.
To quickly get some independent perspective, I recommend reading the Overview of the handbook part of “The Handbook of Rationality” (2021, MIT Press, open access). For an extremely crude calibration: the Handbook has 65 chapters. I’m happy to argue at least half of them cover topics relevant to the EA project. About ~3 are directly about Kahneman and Tversky’s work. So, by this proxy, you would miss about 90% of whats relevant.
Sorry for sarcasm, but what about returning to the same level of non-involvement and non-interaction between EA and Rationality as you describe was happening in Sydney? I.e. EA events are just co-hosted with LW Rationality and Transhumanism, and the level of Rationality idea non-influence is kept on par with Transhumanism?
It would be indeed very strange if people made the distinction, thought about the problem carefully, and advocated for distancing from ‘small r’ rationality in particular.
I would expect real cases to look like
- someone is deciding about an EAGx conference program; a talk on prediction markets sounds subtly Rationality-coded, and is not put on schedule
- someone applies to OP for funding to create rationality training website; this is not funded because making the distinction between Rationality and rationality would require too much nuance
- someone is deciding about what intro level materials to link to; some links to LessWrong are not included
The crux is really what’s at the end of my text—if people do steps like above, and nothing else, they are distancing also from the ‘small r’ thing.
Obviously part of the problem for the separation plan is Rationality and Rationality-adjacent community actually made meaningful progress on rationality and rationality education; a funny example here in the comments … Radical Empath Ismam advocates for the split and suggests EAs should draw from the “scientific skepticism” tradition instead of Bay Rationality. Well, if I take that suggestion seriously, and start looking for what could be good intro materials relevant to the EA project (which “debunking claims about telekinesis” advocacy content probably isn’t) …. I’ll find New York City Skeptics and their podcast, Rationally Speaking. Run by Julia Galef, who also later wrote Scout Mindset. Excellent. And also, co-founded CFAR.
(crossposted from twitter) Main thoughts:
1. Maps pull the territory
2. Beware what maps you summon
Leopold Aschenbrenners series of essays is a fascinating read: there is a ton of locally valid observations and arguments. Lot of the content is the type of stuff mostly discussed in private. Many of the high-level observations are correct.
At the same time, my overall impression is the set of maps sketched pulls toward existential catastrophe, and this is true not only for the ‘this is how things can go wrong’ part, but also for the ‘this is how we solve things’ part.
Leopold is likely aware of the this angle of criticism, and deflects it with ‘this is just realism’ and ‘I don’t wish things were like this, but they most likely are’. I basically don’t buy that claim.
FWIW … in my opinion, retaining the property might have been a more beneficial decision.
Also, I think some people working in the space should not make an update against plans like “have a permanent venue”, but plausibly should make some updates about the “major donors”. My guess this almost certainly means Open Philanthropy, and also likely they had most of the actual power in this decision.
Before delving further, it’s important to outline some potential conflicts of interest and biases:
- I did co-organize or participated at multiple events at Wytham. For example, in 2023, ACS organized a private research retreat aimed at increasing the surface area between Active Inference and AI Alignment communities. The event succeeded to attracts some of the best people from both sides and was pretty valuable for the direction of alignment research I do care about, and the Oxford location was very useful for that. I regret running events like that in future will be more difficult.
- I have friends in all orgs or sides involved—Wytham project, Open Phil, EV, EAs who disapproved the purchase,...
- I lead an org funded by Open Philanthropy—
I also lead an org which was fiscally sponsoring a different project of venue purchase, which was funded by FTX regrant (won’t comment on that for legal reasons)
Also, without more details published, my current opinion is personal speculation, partially based on my reading of the vibes.
My impression-from-a-distance is part of the decision is driven by a factor which I think should not be given undue weight, and a factor where I likely disagree
Factor where I possibly disagree is aesthetics. As far as I can tell, the current preferred EA aesthetics is something more similar to how recent EAG Bay looked like. At EAG Bay, my impression of the vibes of venue was… quite dystopian—the main space is a giant hall with unpleasant artificial lights, no natural light, no colours, endless rows of identical black tables utilized by people having endless rows of 1:1s. In some vague aesthetic space, nearby vibes vectors are faceless bureaucracies, borgs and the scifi portrayals of heart-less technocratic baddies. Also something about naive utilitarianism and army.
Wytham seemed to stand in stark contrast to this aesthetics: the building was old and full of quirks. The vibes were more like an old Oxford college.
Factor which I would guess was part of the decision and I suspect had weight was PR concerns. Wytham definitely got some negative media coverage both in traditional media, social media, and on this forum.
What I dislike about this, these concerns often seemed to be mostly on the Simulacra levels 3 and 4, detached from the reality of running events in Oxford, or actual concern about costs. (Why I do think so? Because of approximately zero amount of negative PR, forum criticisms etc. anyone and anything in the ecosystem is getting for renting properties, even if they are more expensive per day or per person.)
To be clear—
I don’t think these were the only or main(?) factors.
- * I would expect somewhere also exists some spreadsheet with some estimates of “value” of events at Wytham. If this is the case, I probably also disagree about some of the generative opinions about what’s valuable.
Still, given the amount of speculative criticism the purchase of Wytham generated on the forum, it seems good for transparency to also express critical view about the sale.
In my view the basic problem with this analysis is you probably can’t lump all the camps together as one thing and evaluate them together as one entity. Format, structure, leadership and participants seem to have been very different.
Based on public criticisms of their work and also reading some documents about a case where we were deciding whether to admit someone to some event (and they forwarded their communication with CH). It’s a limited evidence, but still some evidence.
This is a bit tangential/meta, but looking at the comment counter makes me want to express gratitude to the Community Health Team at CEA.
I think here we see a ‘practical demonstration’ of the counterfactuals of their work:
- insane amount of attention sucked by this
- the court of public opinions on fora seems basically strictly worse at all relevant dimensions like fairness, respect of privacy or compassion to people involved
As ‘something like this’ would be quite often the counterfactual to CH to trying to deal with stuff …it makes it clear how much value they are creating by dealing with these problems, even if their process is imperfect
Sorry for the delay in response.
Here I look at it from a purely memetic perspective—you can imagine thinking as a self-interested memplex. Note I’m not claiming this is the main useful perspective, or this should be the main perspective to take.
Basically, from this perspective
* the more people think about AI race, the easier is to imagine AI doom. Also the specific artifacts produced by AI race make people more worried—ChatGPT and GPT-4 likely did more for normalizing and spreading worried about AI doom than all the previous AI safety outreach together.
The more the AI race is clear reality people agree on, the more attentional power and brainpower you will get.
* but also from the opposite direction… : one of the central claim of the doom memplex is AI systems will be incredibly powerful in our lifetimes—powerful enough to commit omnicide, take over the world, etc. - and their construction is highly convergent. If you buy into this, and you are certain type of person, you are pulled toward “being in this game”. Subjectively, it’s much better if you—the risk-aware, pro-humanity player—are at the front. Safety concerns of Elon Musk leading to founding of OpenAI likely did more to advance AGI than all advocacy of Kurzweil-type accelerationist until that point...
Empirically, the more people buy into the “single powerful AI systems are incredibly dangerous”, the more attention goes toward work on such system.
Both memeplexes share a decent amount of maps, which tend to work as blueprints or self-fullfilling prophecies for what to aim for.
Personally, I think the 1:1 meme is deeply confused.
A helpful analogy (thanks to Ollie Base) is with nutrition. Imagine someone hearing that “chia seeds are the nutritionally most valuable food, top rated in surveys” … and subsequently deciding to eat just chia seeds, and nothing else!
In my view, sort of obviously, intellectual conference diet consisting just of 1:1s is poor and unhealthy for almost everyone.
I think it could be a helpful response for people who are able to respond to signals of the type “someone who has demonstrably good forecasting skills, is an expert in the field, and works on this long time claims X” by at least re-evaluating if their models make sense and are not missing some important considerations.
If someone is at least able to that, they can for example ask a friendly AI or some other friendly AI and they will tell you, based on conservative estimates and reference classes, that the original claim is likely wrong. They will still miss important considerations—in a way in which typical forecaster would also do—so the results are underestimates.
I think at the level of [some combination of lack of ability to think and motivated reasoning] when people are uninterested in e.g. sanity checking their thinking with AIs, it is not worth the time correcting them. People are wrong on the internet all the time.
(I think the debate was moderately useful—I made an update from this debate & voting patterns, broadly in the direction EA Forum descending to a level of random place on the internet where confused people talk about AI and it is broadly not worth to read or engage. I’m no longer that much active on EAF, but I’ve made some update)