alkjash

• alkjash 12 Dec 2024 13:53 UTC

  3 points

  1

  on: Un­der­stand­ing Shap­ley Values with Venn Diagrams

  I taught game theory at Princeton and wish I’d seen this explanation beforehand, excellent framing.

On “Geeks, MOPs, and So­ciopaths”

Through­put vs. Latency

• alkjash 11 Jan 2024 13:44 UTC

  9 points

  4

  in reply to: Raemon’s comment on: Set­ting the Zero Point

  In the territory, bad event happens [husband hits wife, missile hits child, car hits pedestrian]. There is no confusion about the territory: everyone understands the trajectories of particles that led to the catastrophe. But somehow there is a long and tortuous debate about who is responsible/​to blame [“She was wearing a dark hoodie that night,” “He should have come to a complete stop at the stop sign”, “Why did she jaywalk when the crosswalk was just 10 feet away!”].
  
  The problem is that we mean a bunch of different things simultaneously by blame/​responsibility:

  1. Causality. The actual causal structure of the event. [“If she’d worn a reflective vest this wouldn’t have happened,” “If your left headlight wasn’t broken you’d have seen her.”]

  2. Blame. Who should be punished/​shamed in this situation. This question already branches into a bunch of cruxes about the purpose and effectiveness of punishment.

  3. Responsibility. What is the most effective way of preventing such events in the future? [“If we passed a law that all pedestrians wear reflective vests it would halve incidents like this”, “How about we institute mandatory pedestrian-sighting courses for drivers, and not blame the victim?”]

  People argue about the same event with different causal models, different definitions of blame, and different notions of responsibility, and the conversation collapses. Fill in your own politically-charged example.
  
  Setting the zero point seems to be one “move” in this blame game [if the default is that all drivers take pedestrian-sighting courses, then you’re to blame if you skipped it. if the default is that all pedestrians must wear reflective vests, then you’re to blame if you didn’t wear one.]

• alkjash 6 Dec 2023 21:50 UTC

  11 points

  3

  in reply to: Ben Pace’s comment on: Origi­nal­ity vs. Correctness

  I don’t have a complete reply to this yet, but wanted to clarify if it was not clear that the position in this dialogue was written with the audience (a particularly circumspect broad-map-building audience) in mind. I certainly think that the vast majority of young people outside this community would benefit from spending more time building broad maps of reality before committing to career/​identity/​community choices. So I certainly don’t prescribe giving up entirely.

  ETA: Maybe a useful analogy is that for Amazon shopping I have found doing serious research into products (past looking at purchase volume and average ratings) largely unhelpful. Usually if I read reviews carefully, I end up more confused than anything else as a large list of tail risks and second-order considerations are brought to my attention. Career choice I suspect is similar with much higher stakes.

Origi­nal­ity vs. Correctness

• alkjash 20 Sep 2023 19:55 UTC

  2 points

  in reply to: A Pear With legs’s comment on: Write

  Seeing patterns where there are none is also part of my writing process.

• alkjash 27 Jan 2023 23:48 UTC

  8 points

  0

  on: As­sign­ing Praise and Blame: De­cou­pling Episte­mol­ogy and De­ci­sion Theory

  This paper of mine answers exactly this question (nonconstructively, using the minimax theorem).

• alkjash 26 Jan 2023 18:54 UTC

  2 points

  0

  on: Set­ting the Zero Point

  I feel there is an important thing here but [setting the zero point] is either not the right frame, or a special case of the real thing, [blame and responsibility are often part of the map and not part of the territory] closely related to asymmetric justice and the copenhagen interpretation of ethics.

• alkjash 22 Jan 2023 1:59 UTC

  6 points

  0

  in reply to: Dagon’s comment on: Rock, Paper and Scis­sors: A Game The­ory View

  Afaict, the first simple game is not the prisoner’s dilemma, nor is it zero-sum, nor is the prisoner’s dilemma zero-sum.

• alkjash 17 Jan 2023 20:50 UTC

  3 points

  1

  on: Wis­dom Can­not Be Unzipped

  This is not intended as a criticism in any way, but this post seems to overlap largely with https://​​www.lesswrong.com/​​posts/​​k9dsbn8LZ6tTesDS3/​​sazen.

  [Edit: After looking at the timestamps it looks like that post actually came out after, anyway it might be an helpful alternative perspective on the same phenomenon.]

• alkjash 10 Jan 2023 23:22 UTC

  4 points

  0

  on: Re­view AI Align­ment posts to help figure out how to make a proper AI Align­ment review

  Is it just me or are alignment-related post titles getting longer and longer?

• alkjash 10 Jan 2023 4:07 UTC

  7 points

  on: Cup-Stack­ing Skills (or, Reflex­ive In­vol­un­tary Men­tal Mo­tions)

  This post has a lot of particular charms, but also touches on a generally under-represented subject in LessWrong: the simple power of deliberate practice and competence. The community seems saturated with the kind of thinking that goes [let’s reason about this endeavor from all angles and meta-angles and find the exact cheat code to game reality] at the expense of the simple [git gud scrub]. Of course, gitting gud at reason is one very important aspect of gitting gud in general, but only one aspect.

  The fixation on calibration and correctness in this community trades off heavily against general competence. Being correct is just a very special case of being good at things in general. Part of Duncan’s ethos is that it’s possible to learn [the pattern of gitting gud], and furthermore this is more important and consistent than learning how to be good at one particular arbitrary skill.

• alkjash 1 Jan 2023 19:01 UTC

  7 points

  0

  in reply to: Valentine’s comment on: Slack mat­ters more than any outcome

  It seems important to notice that we don’t have control over when these “shimmying” strategies work, or how. I don’t know the implication of that yet. But it seems awfully important.

  A related move is when applying force to sort of push the adaptive entropy out of a certain subsystem so that that subsystem can untangle some of the entropy. Some kinds of meditation are like this: intentionally clearing the mind and settling the body so that there’s a pocket of calmness in defiance of everything relying on non-calmness, precisely because that creates clarity from which you can meaningfully change things and net decrease adaptive entropy.

  Two further comments:
  (a) The main distinction I wanted to get across is while many behaviors fall under the “addiction from” umbrella, there is a whole spectrum of how more or less productive they are, both on their own terms and with respect to the original root cause.
  (b) I think, but am not sure, I understand what you mean by [let go of the outcome], and my interpretation is different from how the words are received by default. At least for me I cannot actually let go of the outcome psychologically, but what I can do is [expect direct efforts to fail miserably and indirect efforts to be surprisingly fruitful].

  Yeah… for some reason, on this particular point, it always does, no matter how I present it. Then people go on to say things that seem related but importantly aren’t. It’s a detail of how this whole dimension works that I’ve never seen how to communicate without it somehow coming across like an attempt to hijack people. Maybe secretly to me some part of me is trying. But FWIW, hijacking is quite explicitly the opposite of what I want. Alas, spelling that out doesn’t help and sometimes just causes people to say they flat-out don’t believe me. So… here we are.

  Sure, seems like the issue is not a substantive disagreement, but some combination of a rhetorical tic of yours and the topic itself being hard to talk about.

• alkjash 1 Jan 2023 18:03 UTC

  11 points

  5

  on: Slack mat­ters more than any outcome

  Very strongly agree with the part of this post outlining the problem, your definition of “addiction” captures how most people I know spend time (including myself). But I think you’re missing an important piece of the picture. One path (and the path most likely to succeed in my experience) out of these traps is to shimmy towards addictive avoidance behaviors which optimize you out of the hole in a roundabout way. E.g. addictively work out to avoid dealing with relationship issues ⇒ accidentally improve energy levels, confidence, and mood, creating slack to solve relationship issues. E.g. obsessively work on proving theorems to procrastinate on grant applications ⇒ accidentally solve famous problem that renders grant applications trivial.

  And included in this fact is that, best as I can tell, anyone who really groks what I’m talking about will want to prioritize peeling off adaptive entropy over any specific outcome. That using addiction or any other entropy-inducing structure to achieve a goal is the opposite of what they truly want.

  This paragraph raised my alarm bells. There’s a common and “pyramid-schemey” move on LW to say that my particular consideration is upstream and dominant over all other considerations: “AGI ruin is the only bit that matters, drop everything else,” “If you can write Haskell, earning-to-give overwhelms your ability to do good in any other way, forget altruism,” “Persuading other people of important things overwhelms your own ability to do anything, drop your career and learn rhetoric,” and so on ad nauseum.

  To be fair, I agree to a limited extent with all of the statements above, but over the years I’ve acquired so many skills and perspectives (many from yourself Val) that are synergistic and force-multiplying that I’m suspicious any time anyone presents an argument “you must prioritize this particular force-multiplier to the exclusion of all else.”

• alkjash 22 Nov 2022 3:02 UTC

  4 points

  0

  in reply to: Elizabeth’s comment on: Here’s the exit.

  Sure, no big deal.

• alkjash 21 Nov 2022 22:52 UTC

  3 points

  1

  in reply to: Elizabeth’s comment on: Here’s the exit.

  You can’t fight fire with fire, getting out of a tightly wound x-risk trauma spiral involves grounding and building trust in yourself, not being scared into applying the same rigidity in the opposite direction.

  The comment is generally illuminating but this particular sentence seems too snappy and fake-wisdomy to be convincing. Would you mind elaborating?

• alkjash 16 Nov 2022 22:52 UTC

  4 points

  on: Christo­pher Alexan­der’s ar­chi­tec­ture for learning

  Thanks for sharing this, it puts into relief a problem I’ve noticed about academic research: the real research happens behind closed doors and in private communications that the young people don’t have access to. Young people end up only learning about the finished theorems much later on in polished form.

• alkjash 16 Jul 2022 13:37 UTC

  4 points

  in reply to: Venusian’s comment on: Ham­mer­time Day 5: Com­fort Zone Expansion

  That’s great to hear, I’ve been slowly working on this myself in recent years. E.g., it’s greatly improved my gaming experience—from being a total lurker to engaging with Discords, posting bugs and suggestions, occasionally writing Steam guides—it’s enriching for sure.

• alkjash 31 Oct 2021 1:00 UTC

  7 points

  in reply to: MondSemmel’s comment on: Recom­mend­ing Un­der­stand, a Game about Discern­ing the Rules

  I don’t know about skills plural, but the game definitely drilled in that particular skill of aiming to falsify one’s hypotheses instead of just confirming them. That’s a skill well worth a dozen hours of deliberate practice in my opinion.