sarahconstantin

On Try­ing Not To Be Wrong

• sarahconstantin 11 Nov 2016 20:45 UTC

  30 points

  on: Yud­kowsky vs Trump: the nu­clear show­down.

  First of all, I think anecdotal evidence from personal experience with people IRL is always extraordinarily compelling. When, for instance, I go onsite with a client, or when I go to a workshop and do research with people, I feel that I’ve been vastly more productive per hour than I am when I’m on my own. The standard startup advice is always “talk to customers/​users.” Talking face to face with the people who do a thing has outsized power in teaching you about the thing.

  On the other hand, some of this impression can be illusion. Social reality is strong. Being around people IRL might make you feel like you’re learning a lot very fast, but it might just be fairy glamor. Eyewitness accounts are famously unreliable testimony in court.

  Basically, how much is it right to update on “I met some people who actually worked on national security, and I tried doing national-security stuff with them, and believe me, it is very serious and very hard and Trump would fuck it up”? How valuable is Eliezer’s eyewitness testimony?

  I think eyewitness testimony is probably one of the important correctives to news and polls and opinion pieces—data is better than analysis on the margin, especially if you have reason not to trust the analysis, and the analysis is based on limited data. That’s one of my major updates. For all the flaws of eyewitness testimony, I think we need more of it, given the failures of other kinds of data in predicting the election result.

  “I saw this thing. I tried to do this thing. It is very hard. You do not want to elect someone who can’t do the thing.” is a meaningful piece of evidence. If handling foreign policy is a “thinky” kind of activity that takes lots of care and self-restraint, and it seems like Eliezer’s experience is enough to confirm that it is, then we do have evidence from Trump’s lifelong career that he will be bad at that, given that he has a very short attention span and poor impulse control. It’s still possible that he will butt out of anything too technical and leave it to experts.

  “Wars are caused by lack of clarity on which boundaries will be defended” sounds intuitively right, but I know little enough about military history to have really any opinion on that. If the President has a habit of shooting his mouth off, will that cause wars, or will people just stop believing the President? I’m genuinely uncertain.

  Is Eliezer fucking up in a big way by being this anti-Trump? I put a little credence in “Trump is not unusually dangerous and a reasonable person should never have thought he was”, but he is definitely weird and different and this is not a normal election. I don’t buy “the world is in a business-as-usual state with no big trends” as a model to begin with. There are trends, there is “history”, large changes do happen. So, no, I don’t think Eliezer is fucking up in that sense.

  Was it a mistake for him to have a loose, casual FB page with sloppier thinking than his more formal writing? Well, everyone on the internet has moved from more formal to sloppier forms of communication, and I think that is a bad move, and we should all be quite worried about that.

• sarahconstantin 11 Nov 2016 20:54 UTC

  4 points

  on: Did so­cial de­sir­a­bil­ity effects mask Trump’s true sup­port?

  Social desirability is a big one, thanks, I didn’t think of that.

  A related issue is that every representation we see of reality—every newspaper, blog, TV show, movie—is created by a writer or artist, definitionally. So when we look at representations of the world, we’re looking at the biased viewpoint of the highly verbal and well-educated. And, of course, these are not all the people. If you live in a highly symbolic/​representational world, you may be very mistaken about how things work, not only because of a particular bias in your bubble of representations, but because of a systematic bias in all representations.

  (I think Tolstoy is an unusually valuable resource because he writes about the inner lives of the kind of people who would never, ever write or talk about their inner lives.)

  Social desirability bias is a part of this—it biases what we believe towards what people are willing to talk about—but I think the entire phenomenon is worth thinking about.

• sarahconstantin 15 Nov 2016 1:51 UTC

  6 points

  in reply to: baskets’s comment on: On Try­ing Not To Be Wrong

  Point. The polls were not hugely off.

  The felt sense (which I had) that Trump was a ‘crazy, impossible’ candidate and so wouldn’t really win, was off.

  Factual data is useful, and even biased data (e.g. polls, government-published statistics, scientific papers) are useful, and the fact that polls & Nate Silver were not hugely wrong is evidence that we don’t have to throw them out entirely.

  What’s scarily non-useful is opinion.

• sarahconstantin 15 Nov 2016 2:33 UTC

  2 points

  on: Open thread, Nov. 14 - Nov. 20, 2016

  Am I wrong to suspect that genetically engineered microbes could be a boon for fuel or fertilizer on the order of the Haber process? Does anybody know better than I how far along we are on this path?

• sarahconstantin 15 Nov 2016 16:01 UTC

  1 point

  in reply to: CellBioGuy’s comment on: Open thread, Nov. 14 - Nov. 20, 2016

  wait, what am I missing? bacteria already fix nitrogen.

In­dus­try Matters

• sarahconstantin 19 Nov 2016 7:31 UTC

  4 points

  on: Group ra­tio­nal­ity—bridg­ing the gap in a post-truth world

  I think it’s going to have to be a structural change. Encouraging exit from social media (which I’ve just done myself) and building vibrant discussion communities (e.g. increasing engagement at LessWrong itself) seem like good moves.

Vestibu­lar Stim­u­la­tion and Fat Loss

• sarahconstantin 20 Nov 2016 23:13 UTC

  1 point

  on: Dialec­tic al­gorithm—For calcu­lat­ing if an ar­gu­ment is sus­tained or refuted

  This sounds like potentially a mechanic for a discussion site that keeps track of arguments.

• sarahconstantin 22 Nov 2016 21:10 UTC

  2 points

  on: Wire­head­ing Done Right: Stay Pos­i­tive Without Go­ing Insane

  The useful & novel aspect of this seems to be the part about the “transition probability matrix.” For a long time I’ve thought of my emotions as like a Markov chain; it never occurred to me that one could empirically find transition probabilities. This implies that the way to avoid depression is to aim for the “bridge” emotions (like “relief” or “hope”) to make negative states easier to escape.

• sarahconstantin 22 Nov 2016 21:13 UTC

  2 points

  in reply to: DataPacRat’s comment on: Vestibu­lar Stim­u­la­tion and Fat Loss

  I’m going to try to do this myself. It involves a battery, an inverter (to convert it to AC), and electrodes. I’ve been warned that you need gel to avoid pain. I’ll also caution that we don’t know about the effects of prolonged galvanic stimulation on humans, so if you’re doing it for more than an hour at a time, you’re in unexplored waters.

• sarahconstantin 22 Nov 2016 22:37 UTC

  11 points

  on: Less costly signaling

  One main prescription of the article seems to be “encourage signaling of prosocial goods, so that the cost at least goes to somebody.” I think this does not necessarily work, because if something’s value as a signal depends on its cost and not on its prosocial qualities, then the natural pressure will optimize away from real benefits and towards apparent benefits.

  What you get, I think, if you aim for “at least make your signaling do-gooder-ish” is a proliferation of fake do-gooding, which has the real harm of giving actual do-gooding a bad name. I’m not sure that this is net harmful, but I think it’s a matter of real doubt. “Conspicuous philanthropy” has, after all, provided the US with countless Carnegie libraries. But it’s a serious question whether poorly-aimed aid might be net harmful for poor countries. For another example, the Crusades were acts of “conspicuous philanthropy”—European nobles did not profit from them, but committed vast sums to foreign wars out of piety. It’s not at all clear that this was good for the world.

In­dus­try Mat­ters 2: Par­tial Retraction

• sarahconstantin 23 Nov 2016 17:14 UTC

  4 points

  in reply to: WhySpace_duplicate0.9261692129075527’s comment on: Less costly signaling

  This seems believable; but losses compared to what is the question.

  It’s hard to maintain “actually trying” levels of effort to achieve objectives in a large organization, simply because it’s hard to communicate across large numbers of people what the objectives are.

  However, large organizations are still economically efficient for some uses. An economy consisting only of five-person startups would, I’m pretty sure, be poorer than the one we live in today.

  It seems that we trade off “degree of honest goal-alignment” with “returns to scale” to some extent.

  Arguably, monopoly power artificially makes organizations larger than they would naturally be, which makes a higher share of the economy into signaling/​makework than it optimally would be. See https://​​www.minneapolisfed.org/​​publications/​​the-region/​​the-costs-of-monopoly-a-new-view, and, like, all of Gabriel Kolko.

• sarahconstantin 23 Nov 2016 18:08 UTC

  4 points

  on: Po­lariza­tion is the prob­lem, “nor­mal­iza­tion” is the answer

  Normalization seems deeply wrong to me (even if you disagree with polarization.)

  “Normalization” seems to be looking at things relative to the population median. If you’re a middling American, you’re normal, and you should feel normal, and be treated as a normal tribe member.′

  This is a very different viewpoint than looking at things relative to an objective external benchmark, and it means that you can’t track societal drift. If you “normalize”, you won’t notice that, say, societal attitudes on LGBT issues have liberalized a lot very fast, or that the omnipresence of the Internet is societally very weird. You won’t notice when certain issues have outsize objective importance in the number of people they affect (like x-risk or global economic growth.)

  If, indeed, Trump is an unprecedentedly terrible president (I’m not confident he is, but he might be) then it is really counterproductive to “normalize” him, in the sense of saying “this is the new normal, because this is typical of today’s world.” That’s like those sitcoms that go on for twenty years without the characters ever getting older. It’s an illusion of stasis in the midst of rapid objective change.

  The way you would notice that slavery was wrong, in 19th-century America, would be to set your benchmark somewhere outside a mere opinion-poll average of typical Americans. “Normalization” never gives you the chance to personally be more ethical than your generation.

  What links here?

  • In­flated Bubbles by Jacob Falkovich (23 Dec 2016 16:55 UTC; 1 point)

• sarahconstantin 24 Nov 2016 22:05 UTC

  6 points

  in reply to: Benquo’s comment on: In­dus­try Mat­ters 2: Par­tial Retraction

  In autos, at least, we look worse when you measure “number of widgets” than when you measure dollar value. The total number of cars produced in the US has actually decreased. The peak was in 1978.

  (I seem to recall that the US boosted sales by switching to SUVs, which were more expensive per unit.)

• sarahconstantin 24 Nov 2016 22:08 UTC

  6 points

  in reply to: Benquo’s comment on: In­dus­try Mat­ters 2: Par­tial Retraction

  My argument had been that if we were moving up the value chain successfully we’d see higher productivity, and we don’t.

  My intuition is that something to do with knowledge/​skill/​synergy goes away if you move too much manufacturing away, which makes the manufacturing that remains less productive. For example, there are efficiency gains from having factories close to suppliers; if the parts manufacturers move to China, then the auto manufacturers in the US don’t get to have the same tight feedback loops with them, and so become less productive per worker and per dollar of capital, because the benefits of proximity don’t show up in either of those categories.

• sarahconstantin 25 Nov 2016 15:43 UTC

  8 points

  on: Old urine sam­ples from the 2008 and 2012 Olympics show mas­sive cheating

  The takeaway should be an update towards mistrust in institutions.

  Presidents swindle and rape—therefore the Presidency is not sacred. The Red Cross steals—therefore the Red Cross is not sacred. Olympic athletes cheat—therefore the Olympics is not sacred. Many scientific articles are fraudulent—therefore peer review is not sacred.

  This is not the same as saying the institutions should not exist, or are worse than the alternative. But it seems very common to have confusion about the trustworthiness of institutions. On the one hand, to know that there is dishonesty; on the other hand, to continue to trust the institutions as though they were honest. I don’t think this is an example of pure cynicism on the part of ordinary people; I don’t think it means they simply favor the high-status and strong, and truly don’t care about honesty. I think that historically there have been efforts to convince people “you can trust this, it is reliable” about certain things, and that we haven’t quite adapted to knowing we’re no longer in that world. It is very hard to maintain awareness of one’s true aloneness.

  I also think that it’s easy to forget that trust (being able to depend on things, being able to be secure, a world that remains stable and normal and safe) is a function of honesty (having words match exactly to deeds and facts.) We blow off honesty as a “mere” moral scruple, but expect to continue to trust the world around us.

• sarahconstantin 25 Nov 2016 18:52 UTC

  6 points

  in reply to: James_Miller’s comment on: Old urine sam­ples from the 2008 and 2012 Olympics show mas­sive cheating

  I have heard the rumor that drugstore products (e.g. shampoo, wrinkle cream) contain ever-declining amounts of the active ingredient, after the first product launches.