--- title: Notes | Tutorial 38 (Paths Forward) layout: post date: 2020-09-09 11:00 parent: Tutorial Notes tutorial_link: https://www.youtube.com/watch?v=MddhepUvIYw lesson: 38 --- ## Paths Forward * what problem is PF trying to solve * doing communication well is hard and often results in disagreements * when disagreements happen we need ways to resolve them * PF is a process/method/methodology to reliably resolve differences * i.e. people can do a thing and focus on the probelms that exist and thus figure out what to do to sovle them * ET: think of it more specifically * > if i'm wrong, how will I stop being wrong? * and PF particularly deals with: if someone knows I'm wrong and they want to tell me, how can they do that? * you can ask ppl: if I'm right and you're wrong, how will you use this to improve your ideas, will you bet your life on being right? * ET: one reason - ppl already making dozens of bets * another one: I read a bunch of stuff (variety) and try and judge and prioritise them; "your" idea isn't high on the list, I'm still rational just other stuff is more important. * few issue points / potential problems stuff * you could be bad at judging / prioritising * error: not prioritising philosophy stuff (e.g. better at rationality) -> exponential * bad interests * response: "but I think it's important" * triggers other bad responses * we disagree about how important this is; if it is important and you avoid it -> bad long term consequences * need some other rational method to prioritise * ET: specific response - it doesn't answer the question * haven't *actually* offered a path to learning * e.g. what would change their mind * inexplicit -> unaccountable * They've moved it to a meta issue -> topic change to meta topic of how high a priority the issue is e.g dinosaurs (13min) * now there's a new claim it's reasonable to ask the same question: if someone knows I'm wrong and they want to tell me, how can they do that? * also a case - ET *suspects* but they're hiding info * ppl should still want to be corrected bigger issue: how to correct our errors - hard problem. can do it ourself or other ppl might do it. both can be good. ppl have a lack of prep work so other ppl can correct them (e.g. by providing info about your positions, writing, addressing crits, etc) one issue ppl have; they say: * I want to share ideas for crits, * but I only have X hours/energy, * and need privacy * so I'll only share some ideas * ET: * easy case * impersonal stuff only, stuff that's related to an area of speciality/focus * stuff you have a *reason* to know more about * should have paths forward for public intellectual work * excuse above doesn't work * highest value * one can say: if you don't want to write publicly, okay, but don't make claims about it or say ppl are wrong without discussing; be more neutral (21min) * objective state is you withdrew - it's unresolved * hard case common problem: p: I've dealt with this in the past ET: do you have records? p: no, and I don't want to deal with the issue again ET: do you know of other ppls records you agree with p: no 2 big things: 1. if wrong how will you correct that error? 1. what could you say to change their mind? anything? 2. flip side: if you're right, how will *my* error be corrected? 25min written records help (author doesn't matter much) - works both ways * refusing to keep written records is blocking PF for you *and* other ppl * preventing ppl from joining your side ET thinks one big reason is ppl haven't actually read the literature - even on their side. lots of gesturing about things generally; not citing stuff. avoid sticking neck out. makes it harder to engage with. if they're going to say "X author is good" and someone could cherry pick bad ones, why didn't they write down what the bad ones were? (or point to someone else's list) concept "what does it mean for something to contribute to human knowledge" * literature * public, reusable, enduring, sharable (paid is okay; somewhat) * if it's not entered into human knowledge, *you don't know it's right* concept "what counts as entering a thing into human knowledge" * ppl make claims like "I know Rand is bad" but there's no entry for it in human knowledge * no record means no contribution * records need to be * correctable * can correct others * if you haven't done that you haven't added the idea to the stock of human ideas * unfair and unreasonable to bring up ideas that aren't learnable or readable by other ppl * work in progress okay, but don't oversell it * WIP > NOT a WIP; never planning to get to it, etc. * most ppl never plan to get to it ### taking responsibility for ideas 38min * publicly claiming it's true * they're acting like a pub intellectual in some way * partially they're trying to contribute or comment on ideas / project of human knowledge * they should care about what's true and being good at dealing with criticism and ideas * even if not a pub int: you still have ideas that affect your life - learning helps you have a better life; ideas have a big impact * ppl should care about doing a reasonable job of figuring out which ideas are important for them. * what is a pub int? * lots of PPL on LW or reddit are irresponsible about "what if i'm wrong" * "I'm not a PI" -> defense * max's thoughts at 45 min * ET: * ppl don't label their stuff * no disclaimer * relates to dishonesty * lack of disclaimer is to mislead ppl / give wiggle room * ppl *present* themselves as high-status; not much correlation to quality as a thinker tho * disclaimer would be for objective reasons, but they're communicating socially * varies by forum; less excuse on LW than on twitter ET: lack of places where - by default - ppl take responsibility. if you haven't entered a criticism into human knowledge, you don't get to reject it - or one doesn't exist which you can refer to variety of reasons ppl don't want to take responsibility: social climbing, lazy what are the good reasons ppl don't want to take responsibility: "I'm busy" / too much work / too time consuming / other priorities -> big problem -- it takes 0 time if they just *work within their responsibility* 73 min going through "due process" takes too much time 74 min ET: focus helps, not enough of an answer - need rules for getting out of "low quality" bucket - can end up judging ideas by source not content ET: (paraph) debate policy gives freedom and saves time / without it there's pressure 90 min issue: how to deal with volume if you know an idea is low quality, you should have a reason, and that should be written down. - new reason per idea => too much work - crits can refute a *category* of ideas; identify *patterns* - every low qual idea either **fits** existing patterns or is **actually novel** (which is interesting!) - 'actually novel' -> other person can participate, too - library of criticism is like a test suite; can use it to rapidly test new ideas - **todo**: start library of criticism via linking to other indexes of categories of bad ideas - ET himself acts as an index b/c the library is not well organised in all ways that matter; not easy to search everything (thanks google), etc - try to monitor for problems (not supervise) - escalation mechanisms - multiple layers are fine (MK: I think probably necessary to be general/have reach) - person at top is like CEO - needs to take responsibility for *whole* system particularly important point - ppl worry it takes too much time - criticism of CATEGORIES ppl can say "I don't think it fits in cat.A" -> need short bridging argument bridging argument: specifically link idea.X -> cat.A - very short (mostly), but need them frequently bri.args more *general* than other sorts of args? like across populations (even if pops disagree about the correctness of the ideas bridging args reference) -- different sort of reach from normal explanatory reach 107min ppl: "reading a long book is lots of work" ET: same as before: has anyone *else* done the work; if no refutations it means no-one has a good reason (written) about why it's bad. => Big opportunity for the person! big asymmetry between FI and other communities like LW. more ppl on your side who can't written crit responses => be MORE worried about your side having only a few articles between like 1 or 5 ppl is somewhat expected, not with 500k ppl cut out briefly at 113 minutes back at 115 ppl have problem resolving one idea * make some notes about PF, see what I remembered / left out * read PF articles