Harrison Durland

• Harrison Durland 1 Jul 2022 15:21 UTC

  77 points

  0 ∶ 0

  in reply to: John G. Halstead’s comment on: The Fu­ture Might Not Be So Great.

  I don’t think (or, you have not convinced me that) it’s appropriate to use CEA’s actions as strong evidence against Jacy. There are many obvious pragmatic justifications to do so that are only slightly related to the factual basis of the allegations—I.e., even if the allegations are unsubstantiated, the safest option for a large organization like CEA would be to cut ties with him regardless. Furthermore, saying someone has “incentives to lie” about their own defense also feels inappropriate (with some exceptions/​caveats), since that basically applies to almost every situation where someone has been accused. The main thing that you mentioned which seems relevant is his “documented history of lying,” which (I say this in a neutral rather than accusatory way) I haven’t yet seen documentation of.

  Ultimately, these accusations are concerning, but I’m also quite concerned of the idea of throwing around seemingly dubious arguments in service of vilifying someone.

• Harrison Durland 10 Jul 2022 14:51 UTC

  71 points

  0 ∶ 0

  on: Is my blog too provoca­tive for a group or­ga­nizer?

  I’m definitely not the best person to give feedback on this, but I’ll just briefly share a few thoughts:

  1. I’ve heard that EA grant makers often have relatively little time to review grant applications. This may or may not be true for EAIF, but supposing it is, even yellow flags like that offset article might cause a reviewer to quickly become pessimistic about providing grants (for some of the following reasons).

  2. I would have recommended not using the example of rape; murder offsets probably would have been a better alternative. I only skimmed the post, but it really didn’t help that towards the beginning you make the seemingly-intentionally-controversially phrased point about “[Sometimes rape is permissible… you probably agree deep down. That is, if it is to prevent more rape.]” This ordering (saying “you probably agree” before clarifying “if it were in some twisted trolley problem scenario”) and phrasing (e.g., “deep down…”) is needlessly controversy-inviting. To be honest, to me these kinds of details genuinely do reflect some lack of perspective/​room-reading or rhetorical finesse, regardless of whether you ultimately oppose the idea of rape offsets. (It also very much gives me flashbacks to the infamous Robin Hanson post, which really hurt his reputation and reflected a similar lack of perspective…) This may not be such a problem if I am personally evaluating your character, but:

  3. Grant making probably is justified for being cautious about downside risks, including when it comes to optics risks. “EA grant makers fund writer of blog that callously discusses ‘rape offsets’” might be a very unfair social media characterization, but fairness doesn’t really matter, and I can’t be confident it won’t get pulled into some broader narrative attacking—however fairly or unfairly—EA overall. (Speaking as someone who’s never analyzed grant applications) I suspect you would have to have a really good case for potential upside to make it worth spending a few extra hours analyzing those optics risks, and in the end there may (or may not?) be plenty of other people to fund instead.

  As for your overall blog, I haven’t read it, but I wouldn’t be surprised if it is otherwise good, and I’m glad to see a blog discussing moral issues. But rape is a topic that needs to be treated with a lot of care and caution, and probably should be avoided when it is just being used to make a point separate from rape.

• Harrison Durland 21 Aug 2022 16:29 UTC

  40 points

  0 ∶ 0

  on: Should EA shift away (a bit) from elite uni­ver­si­ties?

  As someone who organized at the University of Mississippi—a school of loosely 20K undergrads—I can say that I understand this sentiment and there probably is some merit to it, but I am also somewhat skeptical.

  Tl;dr: Outreach capabilities presumably do not linearly scale with university size, and I suspect higher-prestige universities probably attract a substantially greater proportion (5x-20x?) of engagement-adjusted EA-inclined students than most middle-tier universities, which may offset the bigger pool at larger universities. But there are probably some mid- to high-tier universities with large populations worth targeting, and there may be other meaningful benefits (e.g., diversity)

  TBC, I agree that there are probably a variety of >30K undergrad schools that might be good places to organize at, in part due to the larger population. Additionally, there are probably meaningful benefits from having a more socioeconomically and/​or educationally diverse population (even if only so that EA doesn’t seem as elitist).

  However, I found that despite Ole Miss’ population being larger by a sizable factor than many ivy schools, organizing at Ole Miss was far less effective relative to the experiences I’ve heard from many Ivy League schools.

  With the caveats that I may have simply not been a good organizer (in part due to personality and in part due to the fact that I only had 3 semesters left and became busy with my thesis in the last two), the person who said they’d co-organize with me ended up flaking (and I wasn’t able to find another co-organizer), and COVID hit the semester after it was created: I probably put in some ~5-10 hours a week on average for most of the semester when I began the org, and still struggled to get people interested.

  I put up fliers, blasted groupmes, told people in clubs (especially debate, which probably made up about 50% of regular attendance), tabled at the fresher fair, invited friends, ran informational sessions in welcome week, asked faculty to send emails to departments, got the org included in the honors college newsletter, gave brief presentations in a few classes totaling some 150–200 students (e.g., philosophy and ethics classes), got school funding for free pizza, and at one point even submitted an unrelated article to a student publication on the condition that the editor would attend a discussion meeting (his idea, not mine)—but still failed to attract a co-organizer or maintain sufficient attendance into the second semester.

  (At the time, I was not aware of retreats and fellowships, and probably wouldn’t have had the capacity; they also might have been fairly new/​limited in 2019)

  In contrast, the stories I’ve heard and seen at more prestigious universities tend to be that even with a significantly smaller student population, a number of students happily dive into EA events with relatively little hand-holding or encouragement, and some even happily volunteer to help out with organizing.

  In summary, some of my thoughts have been:

  1. I think that ivy leagues just attract a significantly greater proportion of EA-inclined people. In contrast, I think that many high-impact/​proto-EA/​EA-inclined people just don’t choose to attend most middle-tier universities like Ole Miss. Personally, I was a proto-EA through most of high school and college, and the reasons that I attended Ole Miss are complicated/​uncommon (probably including: I was homeschooled, a bit lazy about college choices, simplistically settled on a career path I figured was high-impact without doing much research, that career path was intelligence analysis which isn’t a common minor/​major, and Ole Miss had fantastic scholarships). At the risk of sounding very pretentious, I’ll say I often felt out of place among my peers at Ole Miss—and retrospectively think I was especially out of place now that I’ve interacted with more EAs (and other people) from higher-prestige universities. If an Ivy League population is ~1/​5th that of a large university but the prevalence of Proto-EAs is >5x that at middle-tier large universities (it might be more like 10–20x), then it’s worth prioritizing Ivies, ceteris paribus.

  2. Your reach does not linearly correlate with the size of the university: you shouldn’t expect to reach 5x as many people at a university 5x the size, in part due to attention competition with other organizations (especially Greek). Additionally, it’s probably easier to attract people you know rather than total randos, and you can’t just know proportionally more people at larger universities.

Re­search + Real­ity Graph­ing to Sup­port AI Policy (and more): Sum­mary of a Frozen Project

• Harrison Durland 1 Jul 2022 17:25 UTC

  34 points

  0 ∶ 0

  in reply to: Harrison Durland’s comment on: The Fu­ture Might Not Be So Great.

  If you’re referring to the same point about his claim to be a cofounder, I did just see that. However, unless I see some additional and/​or more-egregious quotes from Jacy, I have a fairly negative evaluation of your accusation. Perhaps his claim was a bit exaggerative combined with being easily misinterpreted, but it seems he has walked it back. Ultimately, this really does not qualify in my mind as “a history of lying.”

• Harrison Durland 22 Mar 2022 14:16 UTC

  31 points

  0 ∶ 0

  on: A Fo­rum post can be short

  Shortforms may sometimes be good ideas, but it’s important that people recognize that shortforms are much less viewed than normal posts (or at least so I’ve heard and sensed, and I can speak to my own personal engagement which is far less for shortforms).

[Question] How might a herd of in­terns help with AI or biose­cu­rity re­search tasks/​ques­tions?

• Harrison Durland 1 Apr 2022 0:00 UTC

  30 points

  0 ∶ 0

  on: Why 80 000 hours should recom­mend more peo­ple be­come drug lords

  Timezones are important things to remember when reading this post on the evening of March 31st.

  (You/​someone may want to add the appropriate April 1st tag)

• Harrison Durland 8 Mar 2024 17:42 UTC

  29 points

  2 ∶ 0

  on: Har­ri­son D’s Shortform

  Seeing the drama with the NIST AI Safety Institute and Paul Christiano’s appointment and this article about the difficulty of rigorously/​objectively measuring characteristics of generative AI, I figured I’d post my class memo from last October/​November.

  The main point I make is that NIST may not be well suited to creating measurements for complex, multi-dimensional characteristics of language models—and that some people may be overestimating the capabilities of NIST because they don’t recognize how incomparable the Facial Recognition Vendor Test is to this situation of subjective metrics for GenAI and they don’t realize NIST arguably even botched MNIST (which was actually produced by Yann LeCun by recompiling NIST’s datasets). Moreover, government is slow, while AI is fast. Instead, I argue we should consider an alternative model such as federal funding for private/​academic benchmark development (e.g., prize competitions).

  I wasn’t sure if this warranted a full post, especially since it feels a bit late; LMK if you think otherwise!

• Harrison Durland 2 Aug 2022 21:17 UTC

  29 points

  0 ∶ 0

  on: 30 sec­ond ac­tion you could take

  I felt uncomfortable about just taking 30 seconds to choose Will, and did end up spending more time reading some of the other bios, but in the end this seems like a rather silly popularity contest, especially given that it basically doesn’t even describe how you should vote (e.g., “who do you think is the most beneficial ‘thinker’?”). Of course, the fact that it is silly and ought to not be given attention by the public does not mean the public will ignore it, so yeah, I think Will is probably a good choice, and I don’t really have any substantive qualms with voting for him without having read many other profiles.

• Harrison Durland 24 May 2022 5:02 UTC

  25 points

  0 ∶ 0

  on: The EA move­ment’s val­ues are drift­ing. You’re al­lowed to stay put.

  Personally, I dislike the title framing on at least two accounts:

  1. As Vaidehi mentioned, I don’t think that “values” have shifted, but rather implementation/​interpretation of fairly steady values (e.g., truth-seeking/​open-mindedness, impact-maximization) have shifted or at least expanded.

  2. “Drift” has a connotation (if not outright denotation) of “drifting [away] [[from the ideal/​target]]” whereas I think that it’s mainly just fair to say “interpretations are shifting.”

• Harrison Durland 14 May 2022 3:33 UTC

  25 points

  0 ∶ 0

  on: The biggest risk of free-spend­ing EA is not op­tics or epistemics, but grift

  I’m not super motivated+available at the moment to do a full write up/​analysis, but I’m quite skeptical of the idea that the default/​equilibrium in EA would trend towards 100% grift, regardless of whether that is the standard in companies (which I also dispute, although I don’t disagree that as an organization becomes larger self-management becomes increasingly complex—perhaps more complex than can be efficiently handled by humans running on weak ancestral-social hardware).

  It might be plausible that “grift” becomes more of a problem, approaching (say) 25% of spending, but there are a variety of strong horizontal (peer-to-peer) and vertical checks on blatant grift, and at some point if someone wants to just thoroughly scam people it seems like it would be more profitable to do it outside of EA.

  I’d be happy to see someone else do a more thorough response, though.

• Harrison Durland 11 May 2022 20:47 UTC

  25 points

  0 ∶ 0

  on: How I torched my biggest ca­reer op­por­tu­nity so far

  When I read the post title (“torched”), I was expecting to see a story of how you totally screwed something up for no good reason, but unless I missed something I would describe this as more like “passed over” rather than “torched.”

• Harrison Durland 30 May 2023 4:48 UTC

  24 points

  7 ∶ 0

  on: Har­ri­son D’s Shortform

  TL;DR: Someone should probably write a grant to produce a spreadsheet/​dataset of past instances where people claimed a new technology would lead to societal catastrophe, with variables such as “multiple people working on the tech believed it was dangerous.”

  Slightly longer TL;DR: Some AI risk skeptics are mocking people who believe AI could threaten humanity’s existence, saying that many people in the past predicted doom from some new tech. There is seemingly no dataset which lists and evaluates such past instances of “tech doomers.” It seems somewhat ridiculous* to me that nobody has grant-funded a researcher to put together a dataset with variables such as “multiple people working on the technology thought it could be very bad for society.”

  *Low confidence: could totally change my mind

  ———

  I have asked multiple people in the AI safety space if they were aware of any kind of “dataset for past predictions of doom (from new technology)”, but have not encountered such a project. There have been some articles and arguments floating around recently such as “Tech Panics, Generative AI, and the Need for Regulatory Caution”, in which skeptics say we shouldn’t worry about AI x-risk because there are many past cases where people in society made overblown claims that some new technology (e.g., bicycles, electricity) would be disastrous for society.

  While I think it’s right to consider the “outside view” on these kinds of things, I think that most of these claims 1) ignore examples of where there were legitimate reasons to fear the technology (e.g., nuclear weapons, maybe synthetic biology?), and 2) imply the current worries about AI are about as baseless as claims like “electricity will destroy society,” whereas I would argue that the claim “AI x-risk is >1%” stands up quite well against most current scrutiny.

  (These claims also ignore the anthropic argument/​survivor bias—that if they ever were right about doom we wouldn’t be around to observe it—but this is less important.)

  I especially would like to see a dataset that tracks things like “were the people warning of the risks also the people who were building the technology?” More generally, some measurement of “analytical rigor” also seems really important, e.g., “could the claims have stood up to an ounce of contemporary scrutiny (i.e., without the benefit of hindsight)?”

  Absolutely seems worth spending up to $20K to hire researchers to produce such a spreadsheet within the next two-ish months… this could be a critical time period, where people are more receptive to new arguments/​responses…?

• Harrison Durland 11 Aug 2021 4:52 UTC

  24 points

  0 ∶ 0

  in reply to: kokotajlod’s comment on: Out­line of Galef’s “Scout Mind­set”

  I don’t have an extensive list written down, and I’ll emphasize that I’m not saying it’s often best to use a pure soldier mindset. Still, here are some initial thoughts/​examples of where hybrid or soldier-leaning mindsets might be better:

  1. People suffering from depression or similar conditions

  2. People who are not used to using a scout mindset and related tools of rationality (and thus, for example, may not be particularly skilled at applying it—which isn’t to say people should never start learning, but quitting cold turkey may not be a good idea either, especially in combination with other points here)

  3. When you’ve established a “reputation”/​habit of communicating with a soldier mindset, switching to scout-mindset communication may cause people to overfilter your statements (e.g., “Normally this person is really enthusiastic about things, but now they’re being extra uncertain, so they are probably just really skeptical). Also applies for communicating with friends in general that expect you to be supportive (at least when it comes to materially unimportant matters where their being wrong isn’t significant/​bad). Also family: I recently witnessed some major family drama and recognized it would be insanely out of place to express caveats/​thoughts in the way I might normally do so in a scout mindset.

  4. Persuasion: speaking as someone who’s done lots of speech and debate—and is even a major advocate for the light side (truth-seeking/​honest) of the persuasion force—I can say that if you want to be persuasive in some instances, it really helps to commit to the mindset “they are ultimately wrong.” Yes, you need to not be a blind donkey and assume literally everything you say is right—you need to understand how the audience and competitors are thinking and know your side’s weaknesses—but ultimately it helps to believe what you are saying when you actually get up to say what you’ve planned to say.

  5. The rock climbing example (I’ll have to come back to this later and explain some details once I can review the book and my annotations there, I just seem to remember really disliking that example) -- UPDATE: This example is around pages 106–108 in my book. I might have slightly overestimated how much I disliked it when I first read it (based on my minor annotations), but I still think it’s a valid example of where you could get paralyzed by analysis. Of course, you should take some time to consider “what are my alternatives”, but from what I have heard (and just Google-skimmed), climbing down a mountain can often be harder than climbing up—sometimes it’s simply not a choice, unlike what Julia says. Alternatively, staying put might be another bad idea if the elements are threatening. The ultimate point being: yes, do consider alternatives, but if you judge that you have none (or they are all worse), then you may just have to really steel your nerves by adopting a soldier mindset to make the jump without flinching. Julia tries to explicitly respond to this sentiment in the next couple of vignettes/​paragraphs, but I found the responses really lacking for that specific case (e.g., “there’s no clear divide between the ‘decision-making’ and ‘execution’ stages of pursuing a goal” [p. 110]—which definitely seems false for the example of “deciding to make the jump” vs. “making the jump”).

  That’s definitely not an exhaustive list; just initial thoughts/​examples that came to mind. And some of those were addressed in the book, but I feel like the emphasis was much more on “here are why those examples aren’t always justified” than on a more-balanced perspective.

• Harrison Durland 10 Aug 2021 1:57 UTC

  23 points

  0 ∶ 0

  on: Out­line of Galef’s “Scout Mind­set”

  For the majority of people/​instances, I think people would benefit themselves if not also society more broadly (when their goals are not anti-societal) if they would adopt more of a scout mindset in a variety of their decisions. That being said, after reading the book I did at times worry that Julia may have overstated the value or understated the costs/​harms of adopting more of a scout mindset. Or at least, I think there are definitely a non-trivial number of people+decision combinations for which a hybrid or even soldier-heavy mindset is really important, and I felt like the book puts a lot of effort into making the general case for a scout mindset, whereas I had come in hoping for a greater degree of discussion/​analysis on when the scout mindset is appropriate vs. when a more-soldiery mindset is appropriate.

  Overall, I definitely liked the book, and I don’t think this comment does justice to my appreciation of other aspects of the book, but I did want to share that perspective (which some other people also voiced in a book club which covered this book). I think it’s a matter of understanding the audience and having the right expectations about what the book is (a very readable commentary on a broad concept) vs. isn’t meant to be (e.g., an academic study or personal-policy analysis on choosing scout vs. soldier).

  What links here?

  • How should EA nav­i­gate the di­vide be­tween those who pri­ori­tise epistemics vs. so­cial cap­i­tal? by Chris Leong (16 Jan 2023 6:31 UTC; 21 points)

From Lab­o­ra­to­ries to Lan­guage Models: Can AI Sup­port Ri­gor in the Jun­gle of Policy Anal­y­sis? (Linkpost)

• Harrison Durland 25 Mar 2023 19:08 UTC

  22 points

  8 ∶ 0

  on: My at­tempt at ex­plain­ing the case for AI risk in a straight­for­ward way

  FWIW, I just wrote a midterm paper in the format of a memo to the director of OSTP on basically this topic (“The potential growth of AI capabilities and catastrophic risks by 2045”). One of the most frustrating aspects of writing the paper was trying to find externally-credible sources for claims with which I was broadly familiar, rather than links to e.g., EA Forum posts. I think it’s good to see the conceptual explainers, but in the future I would very much hope that a heavily time-constrained but AI-safety-conscious staffer can more easily find credible sources for details and arguments such as the analogy to human evolution, historical examples and analyses of discontinuities in technological progress, the scaling hypothesis and forecasts based heavily on compute trends (please, Open Phil, something other than a Google Doc! Like, can’t we at least get an Arxiv link?), why alignment may be hard, etc.

  I guess my concern is that as the explainer guides proliferate, it can be harder to find the guides that actually emphasize/​provide credible sources… This concern probably doesn’t make new guides net negative, but think it could potentially be mitigated, maybe by having clear, up-front links to other explainers which do provide better sources. (Or if there were some spreadsheet/​list of “here are credible sources for X claim, with multiple variant phrasings of X claim,” that might also be nice...)

  What links here?

  • Plat­form for Pro­ject Spit­bal­ling? (e.g., for AI field build­ing) by Harrison Durland (3 Apr 2023 15:45 UTC; 7 points)

• Harrison Durland 28 Apr 2022 19:45 UTC

  22 points

  1 ∶ 0

  on: My bar­gain with the EA machine

  A perhaps-relevant concept for this discussion: Ikigai

  

• Harrison Durland 11 Jan 2023 20:55 UTC

  21 points

  2 ∶ 2

  on: AGI and the EMH: mar­kets are not ex­pect­ing al­igned or un­al­igned AI in the next 30 years

  I’ve been discussing this concept for some time now, so I’m glad to see some people take a more formal stab at it. However, I must say that I’m overall disappointed with this post. I’ll just lay out a few summary points, and if people are actually still reading this deep into the comments and want to hear more thoughts, I can oblige later:

  1. With the *slight* exception of the “you could be earning alpha” section, it does not really get deep into the causal mechanisms for why you should expect markets to be efficient.

  2. I think this post should have done a better job of aggregating and responding to contrary viewpoints; I feel like the post largely bypassed the key arguments (cruxes) of existing critics and went straight to people who were not familiar with the EMH+AGI debate , especially with all the references to empirical evidence (see next point).
     Granted, “better job” implies that the article did this at all, which I don’t recall it really doing, aside from occasional references to other viewpoints (IIRC).

  3. The empirical sections I thought were decent, but they missed the crux of the debate.

     1. The fact is, we don’t seem to have much of any precedent of this kind of scenario, with some debatable exceptions regarding the Cold War /​ Cuban Missile Crisis—yet the authors didn’t even spend that much time focusing on these examples which seemed to be the most relevant.

     2. Overall, I thought that the empirical sections were not very helpful for the debate, aside from perhaps targeting audiences who are the very early stage of the debate and hastily think “I’ll dismiss EMH in general because of X.”
        (I am normally a big proponent of EMH-style reasoning; it’s not like I and many other people I know that are part of this EMH+AGI debate are saying “EMH has never worked!”^[1]).

  4. One cross-cutting objection off the top of my head is that the people with Special And Justified Knowledge may not be able to profit fast enough to correct the market. [I have read other comments’ responses, and respond to one response in the next point]
     This especially applies to two closely related causal mechanisms of the EMH, profit snowballing and dogpiling: “Suppose you have someone who has better insights than everyone else about some asset. They may not be rich and for various related reasons they are unable to immediately correct the market (i.e., the market is actually temporarily inefficient). However, if they are right/​superior, they either a) can keep profiting over and over again until they become liquid/​rich enough to individually correct the market, and/​or b) other people see that this person is profiting over and over again so they jump in and contribute to market correction.”
     The problem is that it might be the case that there is only one or two cycles for profit with AGI until the world goes crazy, but it could take many years for this strategy to actually profit, during which time the market will be “temporarily” inefficient. If real interest rates don’t rise for 15 years, and only start to rise ~5 years before AGI, the market is inefficient for 15 years because small players can’t profit to fix the situation.

  5. “But those are just two causal mechanisms,” the authors/​defenders hypothetically reply, “and sometimes the market still corrects even without those mechanisms; look at the big short! And there are probably enough AI-conscious investors such that they could alter the market...”

     1. First, I think it’s worthwhile to highlight my view that the debate can unproductively explode at this point because the original authors didn’t (IMO) do a good job of laying out their own causal mechanisms. This forces critics into a game of whack-a-mole filled with delays at the need to comment, wait for responses, address new causal mechanisms, parse out alternate branches of disagreement, etc.
        (However, I think the following subpoint addresses a fairly large part of the debate)

     2. Second, I don’t think that the authors did a good job of differentiating between “sudden surprise takeoffs” (e.g., ~1 year of warning time and ~1/​3rd of people believe this) vs. “forecastable takeoffs” (e.g., ~10 years of warning time and >1/​10th of people believe this). This seems somewhat cruxy in at least one direction—against the authors’ viewpoint. Ultimately, (correct me if I’m wrong) it seems that the authors’ proposed strategy for profit relies on the belief that as you get closer to expected AGI date, more/​richer people will start to agree with your predictions (and still see benefits from getting in on profit): otherwise, prescient investors could believe “AI is very likely to occur around year X, but very few people or institutions will recognize this before X-3 years, such that real interest rates probably won’t change much at all until it’s too late, and when they do change

        1. The counterparty/​non-payment risk may be high;

        2. I prefer a 50% chance of being moderately wealthy for 10 years to a 50% chance of being really rich for ~2 years before I die (with a 50% chance of being poor for 10 years if I bet big and am wrong);

        3. The world might experience chaos which undermines my ability to spend money on things I value, etc.”

     3. Third, I don’t think the claim that “there are probably enough AI-conscious investors...” is supported in this post, and I’m hesitant on this point. I am willing to budge, and this could be a fairly important point if we are in a “forecastable/​slow takeoff” scenario, but I would like to see the post focusing on that leaf of the debate, not trying to recreate the trunk of the debate tree. And again, if we are in a “sudden short timeline” scenario, I suspect that this possibility doesn’t matter all that much.

  1. ^

     Sure, some people may hold this view, but a) I’m skeptical you’ll convince them with this article, and b) you can’t just focus on empirics and then declare victory when there are still many critics who have objections you haven’t directly addressed.

  What links here?

  • Lizka's comment on AGI and the EMH: mar­kets are not ex­pect­ing al­igned or un­al­igned AI in the next 30 years by basil.halperin (13 Jan 2023 16:45 UTC; 46 points)