Harrison Durland

Karma: 1,462

Harrison Durland 20 Nov 2022 15:16 UTC
3 points
3 ∶ 0
on: Take the EA Forum Survey and Help us Improve the Forum
The “Which frequency best describes how often you:” question is poorly designed: there is no middle ground between options like “once a year” and “once a month” (e.g., no “once a quarter”), which basically meant most of my responses to that question were inaccurate.

Harrison Durland 14 Oct 2022 22:39 UTC
2 points
1 ∶ 0
on: The US expands restrictions on AI exports to China. What are the x-risk effects?
This is not an answer to your questions, but I can’t resist pointing out that this post/set of questions seems like a decent illustration of the use case of the project idea I describe here: https://forum.effectivealtruism.org/posts/9RCFq976d9YXBbZyq/research-reality-graphing-to-support-ai-policy-and-more

There’s of course no guarantee that something like the envisioned research graph would have the answers to your questions (it’s hard to predict what users will want, but I think graphs handle this user-uncertainty challenge better than traditional literature reviews) but it might help?

Harrison Durland 10 Oct 2022 14:17 UTC
4 points
in reply to: Charles He’s comment on: Harrison D’s Shortform
- I see what you’re saying now: initially I misread your comment as “It seems [~~good~~] to provide a description of your desired software...” which led me to think you were able to access the article (“it”), which confused me. I also didn’t have problems when I tried the link, but of course it’s now obvious that was because I am the editor. I just fixed that—I’m not even sure how I ended up copying the edit post link anyway.
- FWIW, to address (2) and (1b), I was just trying to ask people if there is any “help me find [or build] a software solution” service (preferably but not necessarily within EA), or if I should just post a full message/question about it on the forum. I was not (yet) trying to ask people to find specific software for me. So, for efficiency’s sake, this post was indeed somewhat low effort—hence the posting via shortform. But including the wrong link and overlooking one important word in your reply are definitely mistakes on my part.

Harrison Durland 10 Oct 2022 4:55 UTC
2 points
in reply to: Charles He’s comment on: Harrison D’s Shortform
Ah, does the link just not work, or are you saying it’s not helpful for finding a sufficient software? I realize now that I meant to say “a partial writeup of the project idea [including the screenshot of the software]”, not a writeup containing a list of the desired features.

Harrison Durland 9 Oct 2022 22:05 UTC
2 points
on: Harrison D’s Shortform
I’ve spent hours today trying to find answers for this, and I’m reaching the point where I think it’s worth throwing out this question just in case someone out there can just solve my problem fairly quickly.

Basically I’m trying to find a platform/software solution for a (~10 person?) AI research project idea I’m doing a writeup for, but I’ve searched for a while without success. I have a list of the features it would need to (or preferably) have, as well as some example platforms/software which have all the necessary features except for cornerstones like “capable of real-time collaboration (like on a Google Doc).”

Basically, I have already done a partial writeup on this project idea, which includes screenshots of one software which seemingly has most of the bare-necessity features except real-time collaboration: https://forum.effectivealtruism.org/posts/9RCFq976d9YXBbZyq/research-reality-graphing-to-support-ai-policy-and-more

Thus, I think it’s now worth asking whether there is any “help me find [or build] a software solution” service (preferably but not necessarily within EA)? Or should I just post a question somewhere (e.g., on the normal EA forum)?

Harrison Durland 9 Oct 2022 14:39 UTC
9 points
5 ∶ 1
on: Warning Shots Probably Wouldn’t Change The Picture Much
To be honest, Nate’s analysis about the hope for government action sometimes comes across as if it’s from someone who never studied political economy or other poli-sci topics, assumed governments would act rational, and then concludes governments are hopeless when that assumption is flawed.

If you started out with the assumption that governments are nationally rational and fast-acting, then yes you obviously need to update away from that in light of the lack of COVID response (and many other examples predating COVID). And yes you do need to have some plausible causal chains for success.

But COVID definitely shouldn’t be assumed as proof of hopelessness (or other pessimistic claims like “Warning Shots Probably Wouldn’t Change The Picture Much”), if only given examples of where government did take sensible or at least strong actions in response to threats/events (sometimes without them even occurring). See for example: 9/11, Y2K, Nunn-Lugar Cooperative Threat Reduction program, the asteroid tracking system, etc. Some of these have even been analyzed in posts here on the EA Forum! (Plausibly also worth learning about MADD: Mothers Against Drunk Driving)

Ultimately, factors such as “Probability of affecting the decision-makers personally,” “existing options for response,” “temporal sharpness of harm,” “good political narratives,” “influential advisors,” etc. matter. Are those things guaranteed for AI? No, but I don’t think this COVID situation is a good/sufficient reason to assume that we can’t act rationally in response to a clear warning shot for AI.

Models of success should reflect this uncertainty clearly in their reasoning/modeling, in case later analysis shows this to be false (i.e., warning shots for AI probably won’t matter).

Harrison Durland 3 Oct 2022 12:39 UTC
6 points
1 ∶ 0
in reply to: Gregory Lewis’s comment on: To those who contributed to Carrick Flynn in Oregon CD-6 - Please help now
Yeah, the language in your comment really resonates with me/my emotions and also gives me a more negative view of the OP, yet I am worried about being overly influenced by 1) the quality of the OP (relative to the legitimacy of the underlying points), and 2) my emotions on this.

Ultimately, I think the second order effects still dominate and warrant someone somewhere giving this request a good think-through separate from emotional reactions:
1. Does not providing some funds to Salinas hurt the chances of future EA candidates or advocacy (especially those who might run for or target the Democratic Party) due to Dem opposition/bitterness (regardless of how legitimate such feelings may be)?
2. Does providing funds to Salinas hurt the chances of future EA candidates or advocacy (especially those who who might run for or target the Republican Party) due to Republicans portraying EA as a “fund blue no matter who” movement (regardless of how legitimate such a label may be)?

Harrison Durland 3 Oct 2022 4:33 UTC
5 points
3 ∶ 0
on: To those who contributed to Carrick Flynn in Oregon CD-6 - Please help now
Wow, opinion on this post went really negative since I last checked. I figured its karma would hang out around 0–15…

Harrison Durland 3 Oct 2022 1:20 UTC
7 points
4 ∶ 0
in reply to: Zach Stein-Perlman’s comment on: To those who contributed to Carrick Flynn in Oregon CD-6 - Please help now
I initially downvoted for many of the same reasons. And tbh, I still don’t like this post, as it does give off big “give us money please” vibes without really justifying some of its key claims.

But ultimately I rescinded the downvote because it (sort of) raises a good point: IF people actually believe Flynn/EA undermined Salinas enough to cost the Dems the election, it might really look bad for EA. This leads me to wonder if it isn’t worth just paying some to offset part of the (supposed) damages.

Personally, I’m reluctant to caving to mud throwing and misrepresenting rhetoric used to extort money out of a good cause. Probably if Salinas came out genuinely in favor of pandemic prevention policies I’d probably be quite supportive of providing funds—but otherwise I’d be iffy on it.

But it’s not my money…

Update: Actually, I’m becoming much more pessimistic about funding Salinas unless she clearly supported pandemic preparedness/prevention, because otherwise it would come across as more partisan (“we’re funding this Democrat because… we were told we hurt this Democrat…”). And appeals to political slogans like “short term business success instead of long term sustainability of our world” is honestly a bit intellectually insulting to me.

Harrison Durland 27 Sep 2022 3:08 UTC
2 points
on: Stress Externalities More in AI Safety Pitches
Ultimately, you should probably tailor messages to your audience, given their understanding, objections/beliefs, values, etc. If you think they understand the phrase “externalities,” I agree, but a sizable number of people in the world do not properly understand the concept.

Overall, I agree that this is probably a good thing to emphasize, but FWIW I think a lot pitches I’ve heard/read do emphasize this insofar as it makes sense to do so, albeit not always with the specific term “externality.”

Harrison Durland 17 Sep 2022 4:52 UTC
6 points
2 ∶ 0
on: Anti War Declaration for a better World
I’m really unclear what the point of this is; to be entirely honest, I don’t think it accomplishes anything besides potentially making you feel better. If it does, that’s fine, but I’m not sure this is the place for such simplistic and emotion-laden declarations.

The reality is that war and conflict is a complicated and unfortunate fact of reality, and it’s not very productive to just spout off random hatred against war. Furthermore, sometimes “peace” might be worse than war. In the Civil War, should the North have simply let the South secede and continue the barbaric practice of slavery? Should the UK have just thrown down their arms when Hitler invaded, or when he was gassing Jewish people?

If you’ve been personally affected by war I can understand that you might have a lot of emotions, but to be brutally honest, the degree of thoughtfulness in this declaration is on par with young children; many people I know have come to understand that the world is not so black and white by middle or at least high school.

Harrison Durland 16 Sep 2022 2:00 UTC
2 points
on: What is neglectedness, actually?
Alternatively, if you’re trying to devise a reliable framework for evaluating decisions, you could just abandon INT in favor of things like COILS (disclaimer: this is my own personal work), which is actually specifically tailored for decision evaluation (rather than “evaluating a cause area to indirectly inform decisions”) and is, in my view, actually a relatively reliable framework.

Harrison Durland 6 Sep 2022 3:23 UTC
5 points
on: The Base Rate of Longtermism Is Bad
If a malaria charity fails to achieve anything, people with malaria can try to complain. But if a longterm-oriented charity fails to achieve anything, the only people who could complain would be time travelers!
I appreciate this concern, but as I explained in this comment here, I think this is not a very strong argument against longtermism. To briefly summarize the ideas I explain in more detail there:
1. Many people are already not held very accountable in the near term, despite what we might hope.
2. Near-term interventions can also prove to be relatively unimportant from a long-term lens.
3. You can definitely be held accountable or feel guilty if it becomes apparent in the near-term that your arguments/proposals will actually be bad in the long-term.
4. Due to the massive expected value, Long-termism can probably still just bite the bullet here even if you mostly dismiss the previous points.

Harrison Durland 4 Sep 2022 15:12 UTC
3 points
in reply to: JulianR’s comment on: How to do theoretical research, a personal perspective
I’m a bit confused by your distinction: the question “Did […]

If you can’t find reliable data, that just makes it hard, not theoretical.

The use of “did” vs “would” wasn’t very intentional or precise.

As to the empirical vs. theoretical nature of my hypothesis, it is indeed claiming that certain relationships empirically existed (and, with a lot of caveats, may continue into the future). However, my point was that the research methods I used were much more “theoretical”: I couldn’t do a large-N empirical analysis or controlled experiments to even establish meaningfully-controlled correlation (let alone causation) between the dependent, control, and independent variables, and instead had to rely on lines of reasoning such as:
1. Hypothetical scenarios (e.g., imagine comparing an ambush where both parties have machine guns vs. one where neither side has machine guns)—which is impractical to clinically/experimentally test (I.e., with high reality fidelity)
2. More-qualitative (and somewhat subjective) comparison of case studies, using a large amount of argumentation/theoretical reasoning to deal with the many gaps and flaws in the case comparison (given that, as I noted, there didn’t seem to be any good case comparison pair in the historical record)
3. Agreement with existing theoretical and/or empirical concepts in the literature, such as Biddle’s Modern System.
it basically just seems to say “think about the problem until you can figure out how to test it with traditional empirical methods.”

Well, yeah, what else would you expect? The post describes how you might use argument clashes and oversimplified simulations in thinking about the problem.

Again, perhaps I was being a bit too imprecise with my language? My point is that for some questions (arguably including my thesis), theoretical argumentation has to bear a lot of the analytical burden. This analytical burden can include things like:
1. Explaining why variables Q, K, and W—none of which you could experimentally control for—probably do or don’t affect the relationship;
2. Explaining why your very limited sample size can probably be extrapolated to some other cases;
3. Explaining why some metric is probably a decent proxy for what you actually are trying to measure;
4. Reasoning about hypothetical scenarios which will not actually empirically occur.
(Caveat: all of those activities can be supported by direct reference to supporting data in some situations, but not always.)

In contrast, it seems that much of the “theoretical” research methods described in this post are basically just “use lots of thinking to figure out how to test this empirically against data [at which point these empirical methods do almost all the legwork.]”

There is perhaps some debate to be had over the meaning of “theoretical” research methods: do mathematical proofs or algorithms count as theory? While I’m not universally opposed to using the term in such a context, I think it is much less helpful to use the term “theory” when you’re trying to juxtapose it with empirical methods. This especially feels true if a major reason you support a mathematical proof or algorithm is based on your determination that “this empirically works every single time.” When teaching research methods, I think it’s important to emphasize the differences that I described previously (e.g., legibility/transparency, reliability/consistency, reputation stake) which, in my view, have tended to make empirical methods so much more effective when they can be used.

Harrison Durland 3 Sep 2022 15:41 UTC
7 points
in reply to: Michael_PJ’s comment on: How to do theoretical research, a personal perspective
If that was the intention, I think the title and content should have more clearly expressed that. I’m unclear on what the significant difference is between theory and empirics in math; I think most of the value in distinguishing between theoretical and empirical research comes from highlighting the inability to simply “use data or counter-examples to falsify/test a hypothesis”—but in this case, that doesn’t apply.

Harrison Durland 29 Aug 2022 5:23 UTC
2 points
in reply to: Jack Malde’s comment on: Does this solve Pascal’s Muggings?

It’s worth at least noting that many people wouldn’t do the same

I don’t think many people are capable of actually internalizing all of the relevant assumptions that in real life would be totally unreasonable, nor do most people have a really good sense of why they have certain intuitions in the first place. So, it’s not particularly surprising/interesting that people would have very different views on this question.

Harrison Durland 28 Aug 2022 19:50 UTC
2 points
in reply to: Jack Malde’s comment on: Does this solve Pascal’s Muggings?
Well, with those extra assumptions I would no longer consider it a Pascalian Mugging, I would probably just consider it a prone-to-mislead thought experiment.
Would I take a 1*10 $^{- 10}$ chance of getting 10 $^{15}$ utilons over the option of a 100% chance of 1 utilon? Well, if we assume that:
- Utilons are the only thing that matter for morality/wellbeing (ignore all moral uncertainty),
- I don’t have any additional time to reduce my uncertainty (or I have maximal certainty of the 1*10 $^{- 10}$ chance estimate),
- The 1*10 $^{- 10}$ chance estimate already takes into account all of the outside views, intuitions, and any other arguments that says “this is extremely improbable”,
- There’s no such thing as opportunity cost,
- Diminishing marginal returns don’t apply,
- Psychological and social factors are non-existent,
- Your answer to this hypothetical question will not be used to judge your intelligence/character in the real world,
- (And how about for good measure, “Assume away any other argument which could dispute the intended conclusion”),
Then the answer is probably yes?
I have a hard-to-articulate set of thoughts on what makes some thought experiments valuable vs. misleading, with one of them being something along the lines of “It’s unhelpful to generate an unrealistic thought experiment and then use the answer/response produced by a framework/system as an argument against using that framework/system in the real world if the argument is (at least implicitly) ‘Wow, look at what an unreasonable answer this could produce for real world problems,’ especially given how people may be prone to misinterpret such evidence (especially due to framing effects) for frameworks/systems that they are unfamiliar with or are already biased against.”
But I don’t have time to get into that can of worms, unfortunately.

Harrison Durland 28 Aug 2022 19:07 UTC
2 points
in reply to: Jack Malde’s comment on: Does this solve Pascal’s Muggings?
For example, you can calculate the EV of the possibility of meeting a real wizard in a year and them generating an insane amount of utility. Your EV calculation will spit out some sort of number right? Well the mugger in front of you can beat whatever number that is by saying whatever number they need to to beat it.
Hence why I wrote out “infinite utilons” with such emphasis and in a previous comment also wrote “If we want to just end this back and forth, we’ll just jump straight to claims of infinity [...]”. But you do continue:
Or maybe the EV is undefined because you start talking about infinite utility (as you allude to). In this case EV reasoning pretty much breaks down.
I disagree (at least insofar as I conceptualize EV and if you are just saying “you can’t compare fractional infinities”), as I already asserted:
a 20% chance of achieving an infinitely good outcome has higher expected value (or, is a better option) than a 10% chance of achieving an infinitely good outcome, ceteris paribus.
Summarizing your thoughts, you offer the following conclusion:
Which brings me to the point that rejecting EV reasoning in the first place can get you out of handing over the wallet. Maybe there’s a problem with EV reasoning, for example when dealing with infinite utilities, or a problem with EV reasoning when probabilities get really small.
Indeed I’m not sure we should actually hand over the wallet because I’m open to the possibility EV reasoning is flawed, at least in this case. See the St. Petersburg Paradox and the Pasadena Game for potential issues with EV reasoning.
I’m also somewhat open to the idea that EV reasoning is flawed, especially my interpretation of the concept. However:
1. I am very skeptical of rejecting EV reasoning without any kind of replacement, which I don’t think I’ve seen you offer.
2. This conversation has not moved me closer to believing that I should reject EV reasoning due to Pascalian mugging objections, of which I don’t see any standing/compelling examples. In fact, it is through EV reasoning that I find these objections uncompelling.
3. The St. Petersburg Paradox and Pasadena Game are interesting objections with which I’m decently familiar (especially the former); I’ve written one or two somewhat long comments about the former some time ago. However, these are also not persuasive, for some of the same reasons I’m articulating here (e.g., it might be equally likely that shoving your $100 bill in some random tree hollow is going to reveal a Korok that breaks the Matrix and gives you infinite utilons than that you are going to flip heads for an infinitely long time).

Harrison Durland 28 Aug 2022 17:58 UTC
2 points
in reply to: Jack Malde’s comment on: Does this solve Pascal’s Muggings?
But do you actually have a good reason to believe the costs and opportunity costs will rapidly increase to unimaginably high levels? Why would this be the case?
A few responses:
1. One which I already noted:
“insofar as there is any credence to assign to what this person is saying, there are better reasons to think that sacrificing my resources, reputation, psychological wellbeing, time, etc. will have worse expected value: if it’s actually possible to break the laws of physics and produce a googolplex utilons, it’s unlikely that giving over the wallet is the [best] way to achieve that [in expectation]… etc.”
In other words, there are reasons to believe:
1a. The opportunity costs^[1] are correlated with outcome magnitudes that this person is claiming;
1b. The risk-costs^[2] are correlated (as above...)
2. No, I don’t have that many “good” reasons, but that’s been a key point I’ve tried emphasizing: you don’t need good, explicit reasoning to fight completely absurd reasoning, you just need better/less-bad reasoning, which can still be fairly dubious.
1. ^
  Such as “if I give my wallet over now, I won’t have money to give to a REAL Trass or Smiggledorfian fart goblin later on,” or “if I waste my time and money here, I have less time to ponder the design of the universe and find a flaw in the matrix that allows me to escape and live forever in paradise (or to figure out which supernatural/religious belief is valid and act on that to get infinite utilons).”
2. ^
  Such as, “Actually, this person is testing me, and will do the exact opposite of what they claim if and only if I do what they are telling me to do.”

Harrison Durland 28 Aug 2022 17:37 UTC
2 points
in reply to: Jack Malde’s comment on: Does this solve Pascal’s Muggings?
In the original Pascal’s mugging the key idea is that the mugger can promise more and more utility to overcome any barriers you might have to handing over the wallet. Yes the probability you think the mugger is telling the truth might decrease too, but at some point the probability will stop decreasing or at least won’t decrease as fast as the mugger can promise more.
You’re only responding to the first (and admittedly less important) half of my response. The key issue is that you can just use that same logic against the new claims and still defeat the : for any given increase in the mugger’s claims, you can come up with an equally large (but still more probable) increase in opportunity cost or other objections. If the mugger says “no wait, actually I’ll give you ten googolplex of utilons instead of just one googolplex, and I’ll bet you don’t think that less than 10% as likely,” I could just say “Hmm, yeah maybe it’s only 20% as likely so that would double the expected value, except actually I now think the opportunity costs and other downsides are two to three times greater than what I thought before, so actually the costs still outweigh the gains.”
If we want to just end this back and forth, we’ll just jump straight to claims of infinity: a 20% chance of achieving an infinitely good outcome has higher expected value (or, is a better option) than a 10% chance of achieving an infinitely good outcome, ceteris paribus.
Why can’t I just copy that reasoning and think “by choosing to not hand over my wallet to Trass, the Tin-and-Brass God of Mugging, I am working towards reducing the likelihood of existential catastrophe, which means I am more likely to be reducing it. QED”
Not sure what you’re getting at. I’m saying that, on average, people trying to achieve a goal will mean they are more likely to achieve it than if they weren’t trying to achieve the goal. This was simply to counter Sanjay’s point about AI and is indeed detached to the discussion of Pascal’s mugging.
I don’t think I originally saw that part of Sanjay’s comment/post, and I agree that there is probably something wrong with his arguments there. I partially thought you were making that point as an additional argument in favor of giving your wallet to the mugger, since you would be doing so in order to try to reduce x-risk.