X-Git-Url: http://unremediatedgender.space/source?a=blobdiff_plain;ds=sidebyside;f=notes%2Fmemoir-sections.md;h=a2831d1d1136cecc9754a6563a2c0b99dac54b8a;hb=HEAD;hp=f067e327d9a07b1cc6cecb9aa1aa83d6afb70d12;hpb=7bc920b52eb3640b7e23b3109ed2b9c24cfd8d27;p=Ultimately_Untrue_Thought.git diff --git a/notes/memoir-sections.md b/notes/memoir-sections.md index f067e32..379869a 100644 --- a/notes/memoir-sections.md +++ b/notes/memoir-sections.md @@ -1,35 +1,5 @@ -pt. 4 edit tier— -✓ "A Fire" § title -✓ make sure I'm summarizing "policy debates" moral from "Challenges" -- revise "too good a writer" to be more explicit "someone could be that naive" -_ footnote about how I could be blamed for being too credulous? -_ say that explicitly, up front, at the start of that … chunk. -_ edit post to clarify "nudging the cognition" -_ Tail's objection to FFS example -_ Brennan "everyone else should participate" needs more wording adjustments -_ the mailing list post noted it as a "common sexual fantasy" -_ Sept. 2020 clarification noted that a distinction should be made between -_ emphasize that 2018 thread was policing TERF-like pronoun usage, not just disapproving of gender-based pronouns -_ look for a place to link http://benjaminrosshoffman.com/discursive-warfare-and-faction-formation/ -_ cite more sneers; use a footnote to pack in as many as possible -_ Stephen Jay Gould -_ Dawkins and Jerry Coyne and https://www.thefp.com/p/carole-hooven-why-i-left-harvard -_ parenthetical defending literal fraud? - -time-sensitive globals TODOs— -✓ consult Said -✓ patriate-links script TODOs -- remaining pt. 4 edit tier -- draft #drama strategy opening remarks -_ consult Anna -_ #drama strategy session -_ draft Twitter thread -_ consult lc? +_ comment to lc _ bully Jeff Ladish -_ PUBLISH pt. 4!! - --------- - _ finish pt. 5 _ address auto edit tier to pt. 5 _ solicit red team pt. 5 @@ -39,7 +9,6 @@ _ apply pro edit pt. 5 ✓ consult Tail - _ consult David Xu _ psychiatric disaster private doc @@ -1361,11 +1330,19 @@ It even leaked into Big Yud!!! "Counterfactuals were made for humanity, not huma Still citing it (13 Feb 24): https://www.lesswrong.com/posts/kSq5qiafd6SqQoJWv/technologies-and-terminology-ai-isn-t-software-it-s-deepware +Still citing it (22 Feb 24): https://twitter.com/mlbaggins/status/1760710932047577282 + At least I don't have to link the rebuttal myself every time: https://www.datasecretslox.com/index.php/topic,1553.msg38755.html https://old.reddit.com/r/slatestarcodex/comments/10vx6gk/the_categories_were_made_for_man_not_man_for_the/j7k8fjc/ https://twitter.com/niplav_site/status/1744304380503904616 +Still citing it (8 October 2024): https://x.com/tinkady2/status/1843686002977910799 + +Still citing it (AT THE GODDAMNED SEQUENCES READING GROUP, 15 October): https://www.lesswrong.com/events/ft2t5zomq5ju4spGm/lighthaven-sequences-reading-group-6-tuesday-10-15 + +Still citing it (7 Jan 25): https://www.lesswrong.com/posts/X6soKvPmCJ2fkw5zY/don-t-fall-for-ontology-pyramid-schemes + ------ If you _have_ intent-to-inform and occasionally end up using your megaphone to say false things (out of sloppiness or motivated reasoning in the passion of the moment), it's actually not that big of a deal, as long as you're willing to acknowledge corrections. (It helps if you have critics who personally hate your guts and therefore have a motive to catch you making errors, and a discerning audience who will only reward the critics for finding real errors and not fake errors.) In the long run, the errors cancel out. @@ -1582,6 +1559,8 @@ Arthur Schopenhauer's Die Kunst, Recht zu behalten https://wiki.obormot.net/Arch > If human nature were not base, but thoroughly honourable, we should in every debate have no other aim than the discovery of truth; we should not in the least care whether the truth proved to be in favour of the opinion which we had begun by expressing, or of the opinion of our adversary. https://www.greaterwrong.com/posts/ExssKjAaXEEYcnzPd/conversational-cultures-combat-vs-nurture-v2/comment/4qJk3BZ2oQtFskYLh +William Clifford on "The Ethics of Belief": https://statmodeling.stat.columbia.edu/wp-content/uploads/2023/12/Clifford_ethics.pdf + ----- I shall be happy to look bad for _what I actually am_. (If _telling the truth_ about what I've been obsessively preoccupied with all year makes you dislike me, then you probably _should_ dislike me. If you were to approve of me on the basis of _factually inaccurate beliefs_, then the thing of which you approve, wouldn't be _me_.) @@ -2071,6 +2050,7 @@ https://www.greaterwrong.com/posts/QZs4vkC7cbyjL9XA9/changing-emotions/comment/p https://www.greaterwrong.com/posts/WijMw9WkcafmCFgj4/do-scientists-already-know-this-stuff/comment/EFg4p7KRbthgH3SnR https://www.greaterwrong.com/posts/r5H6YCmnn8DMtBtxt/you-are-a-brain/comment/CsqYAeCPnsAsdQb7d https://www.greaterwrong.com/posts/QB9eXzzQWBhq9YuB8/rationalizing-and-sitting-bolt-upright-in-alarm/comment/2FfTFrA7iKK6kxqMq +https://www.greaterwrong.com/posts/8yCXeafJo67tYe5L4/and-all-the-shoggoths-merely-players/comment/AsedCtPfKAef7DdHw I should have said: _You_ are a product of your time. _Eliezer_ has no integrity. Or rather, he's forgotten that you can't define the word "integrity" any way you want. @@ -2707,9 +2687,6 @@ Scott on puberty blockers, dreadful: https://astralcodexten.substack.com/p/highl https://jdpressman.com/2023/08/28/agi-ruin-and-the-road-to-iconoclasm.html -https://www.lesswrong.com/posts/BahoNzY2pzSeM2Dtk/beware-of-stephen-j-gould -> there comes a point in self-deception where it becomes morally indistinguishable from lying. Consistently self-serving scientific "error", in the face of repeated correction and without informing others of the criticism, blends over into scientific fraud. - https://time.com/collection/time100-ai/6309037/eliezer-yudkowsky/ > "I expected to be a tiny voice shouting into the void, and people listened instead. So I doubled down on that." @@ -2835,3 +2812,570 @@ https://twitter.com/ESYudkowsky/status/1755624226550387013 > "Study science, not just me!" is probably the most important piece of advice Ayn Rand should've given her followers and didn't. There's no one human being who ever lived, whose shoulders were broad enough to bear all the weight of a true science with many contributors. https://www.lesswrong.com/posts/96TBXaHwLbFyeAxrg/guardians-of-ayn-rand + +He's still dunking instead of engaging— +https://twitter.com/ESYudkowsky/status/1760701916739194949 +> Every time I've raised an inscrutable alien baby to hyperintelligence by giving it shots of heroin whenever it correctly predicts the exact next word spoken by fictional good characters, it's learned to be a genuinely good person inside! + + +----- + +> I recently advised somebody to distinguish firmly in her mind between "X is actually true" and "X is the politic thing to say"; I advised drawing a great line and the creation of separate mental buckets. The words you write, taken at face value, seem to be missing some... + +https://twitter.com/ESYudkowsky/status/1356493665988829186 +> ...similar distinctions. There's a distinction between honesty in the form of blurting out the whole truth, and honesty in the form of not uttering lies, and a related thing that's not making public confusions *worse* even if you aren't trying to unravel them. There's... + +https://twitter.com/ESYudkowsky/status/1356493883094441984 +> ...being honest in the privacy of your own mind, and being honest with your friends, and being honest in public on the Internet, and even if these things are not perfectly uncorrelated, they are also not the same. Seeking truth is the first one. It's strange and disingenuous... + +https://twitter.com/ESYudkowsky/status/1356494097511370752 +> ...to pretend that the master truthseekers of any age of history, must all have been blurting out everything they knew in public, at all times, on pain of not possibly being able to retain their Art otherwise. I doubt Richard Feynman was like that. More likely is that, say, ... + +https://twitter.com/ESYudkowsky/status/1356494399945854976 +> ...he tried to avoid telling outright lies or making public confusions worse, but mainly got by on having a much-sharper-than-average dividing line in his mine between peer pressure against saying something, and that thing being *false*. That's definitely most of how I do it. + +----- + +https://twitter.com/ESYudkowsky/status/1076155800144363520 +> I got a lot further as an adolescent on "Ask what a superintelligence would think". Eventually I used that up and have now moved on to "What would Anna Salamon think?" + +https://twitter.com/patio11/status/1766115590429618347 +> Please note that this is extremely, extremely true, and if you follow that to its logical conclusion, certain blogs are on the org chart of e.g. the U.S. in the same haha but absolutely serious way the NYT editorial page is. +(quote-Tweeting someone noting Demis Hassabis linking to Slate Star Codex) + +---- + +https://twitter.com/ESYudkowsky/status/1766144388890243108 +> Some social media company finally went and added easy phone calls to Congress. Personally, I was a little "how noneliezeran!" that Google didn't do this way earlier, or Uber under Travis. But I can see how this would be too scary in the USA, such that China did it first. + +This is kind of scummy behavior! Interesting that he calls scummy consequentialism eliezerian + +----- + +https://twitter.com/ESYudkowsky/status/1767276710041686076 +> What about optimising for finding and error-correcting invalid arguments? + +> In my own experience those are the same mental skill, but perhaps your mileage varies. But at any rate, sure. + +https://discord.com/channels/401181628015050773/458419017602826260/1212913739196669963 + +------- + +Unfortunately, there's still some remaining tension here insofar as the guy continues to lean on "you gotta trust me, bro; I'm from dath ilan and therefore racially superior to you" personality-cult-leader intimidation tactics, which I consider myself to have a selfish interest in showing to third parties to be unfounded. + +With anyone else in the world, I'm happy to let an argument drop after it's been stalemated at 20K words, because no one else in the world is making a morally fraudulent claim to be a general-purpose epistemic authority that has a shot at fooling people like me. (_E.g._, Scott Alexander is very explicit about just being a guy with a blog; Scott does not actively try to discourage people from thinking for themselves.) + +New example from today: a claim that MIRI is constrained by the need to hire people who make only valid arguments, and (in response to a commenter) that in his experience, finding and error-correcting invalid arguments is the same mental skill. + +But elsewhere, this _motherfucker_ has been completely shameless about refusing to acknowledge counterarguments that would be politically inconvenient for him to acknowledge! + +[] + +(Screenshot took place in a publicly-linked server and is therefore OK to share) + +My heart racing, it's tempting to leave a Twitter reply saying, "Actually, in my exhaustively documented experience, you don't give a shit about error-correcting invalid arguments when that would be politically inconvenient for you" + +But ... what good would that do, at this point? As I wrote in the memoir, "We've already seen from his behavior that he doesn't give a shit what people like me think of his intellectual integrity. Why would that change?" + +The function of getting the Whole Dumb Story written down that I was supposed to _move on_. I have _other things to do_. + +--------- + + Oli Habryka gets it! () + Vaniver gets it! () + +Eliezer Yudkowsky either doesn't get it, or is pretending not to get it. I almost suspect it's the first one, which is far worse + +https://www.lesswrong.com/posts/CpvyhFy9WvCNsifkY/discussion-with-eliezer-yudkowsky-on-agi-interventions +> Various people who work or worked for MIRI came up with some actually-useful notions here and there, like Jessica Taylor's expected utility quantilization. + +https://twitter.com/ESYudkowsky/status/1301958048911560704 +> That is: they had to impose a (new) quantitative form of "conservatism" in my terminology, producing only results similar (low KL divergence) to things already seen, in order to get human-valued output. They didn't directly optimize for the learned reward function! + +----- + +Metzger is being reasonable here + +https://twitter.com/perrymetzger/status/1773340617671667713 +> That's a fairly inaccurate way of putting it. It wasn't "poked with a stick", what happened was that gradient descent was used to create a function approximator that came as close as possible to matching the inputs and outputs. It's not like someone beat a conscious entity until it deceptively did what its masters demanded but it secretly wants to do something else; in fact, the process didn't even involve GPT-4 itself, it was the process that *created* the weights of GPT-4. + + +https://twitter.com/ESYudkowsky/status/1433206619135823878 +> I don't know if that quite answers your question - but my answer to a lot of "Well, what about *this* kind of AI?" is "Well, what about it? What does one do with that kind of AI, that stops the world from ending 6 months later when some other kind of AI builds nanotech?" + +https://twitter.com/ESYudkowsky/status/1433207431627042819 +> For concreteness: One can see how AlphaFold 2 is working up towards world-ending capability. If you ask how you could integrate an AF2 setup with GPT-3 style human imitation, to embody the human desire for proteins that do nice things... the answer is roughly "Lol, what? No." + +https://twitter.com/ESYudkowsky/status/1775233618035736819 +(in response to Jessica saying that LLM pretraining favors short term) +> True for now, at least. + +He didn't already know this from the discussion when the Sleeper Agents paper came out? +https://twitter.com/ESYudkowsky/status/1778938096123359256 + +https://twitter.com/QiaochuYuan/status/1726860200421085589 +> i was on a 3-month trial for this position, and the terms of the trial were: i would not, at any point, get to field an actual question from the researchers, because all of their actual questions were *too sensitive* for me to know about + +----- + +https://twitter.com/zackmdavis/status/1786284027126968396 +I spent a whole day (3 May 2024) being anxious after replying to Jim's taking dath ilan seriously—whether or not commenting was a good idea (I could have just let it go, rather than bringing up something that I'd already commented on in the chatroom and might write an essay about), being trash for an entire day doesn't make sense—I should be over it by now. + +Arguments for why it was okay to post (in contrast to the intuition that it was a violation of my foreign policy): I have a legitimate interest in de-cultifying Jim; I said "I haven't prioritized finishing my essay about this", suggesting that the Tweets are a substitute for a post—if I had written a post, a link does seem like it would have been contextually appropriate there; the Eliezerfic fight was in December 2022; bringing it up again 16 months later (which it happens to be contextually relevant) isn't obnoxiously bringing it up all the time (the Dec. 2022 discussion was much more salient to me than to him, so a 16 month reminder might actually be useful for reinforcing the message that he shouldn't try to get away with his "I'm from dath ilan and therefore racially superior to you" bullshit) + +---- + +> I'm pretty sure it's a status move. They hope by holding Nora, Quintin et al in contempt that they'll somehow go away. +https://x.com/jd_pressman/status/1731116301182202084 + +https://www.writingruxandrabio.com/p/the-cost-of-freaking-out-about-things + +---- + +He retweeted— + +https://x.com/ryxcommar/status/1809788144511766556 +> i'm not even exaggerating, benignly faked content is so dangerous. every time you read or watch something that's fake, no matter how low stakes, and you don't take the active step of acknowledging it as such, you are dulling the blade that is your mind. + +Garett Jones on J. D. Vance knowing what he's doing— +https://x.com/GarettJones/status/1813255141316776080 +> Yale Law grad. +> +> He knows what he's doing. +> +> The smarter you are, the greater the moral responsibility to avoid lies and half-truths. + +https://x.com/ESYudkowsky/status/1830393607209288070 +> I cannot easily occupy the frame of mind where you find this very surprising and worthy of a great update! I do not understand why you think my worldview was supposed to strongly predict something else! We are learning something about the difficulty of getting a powerful model to predict a thing that is similar to other things it predicts quite well! + + +He retweeted someone's screenshot of Moldbug (https://x.com/shakoistsLog/status/1831083720104972491) in Sept. 2024, which makes his earlier disavowals look motivated + +September 2024— +> One may believe kids own themselves, and this ruling seems consistent with that. +https://x.com/ESYudkowsky/status/1835133943710130395 + + +Distinguish between deception-in-roleplay vs. the shoggoth having its own goals? + +https://x.com/ESYudkowsky/status/1836511740634501275 +> During the original capabilities testing of GPT-4, when it was hiring a human Taskrabbit to solve a CAPTCHA, GPT-4 (deliberately, we know via CoT) lied to the human, saying it had a vision impairment. (Successfully.) +> +> There's your warning. So now what do we all do about it? + +trolly problem infra-alignment: +https://x.com/ESYudkowsky/status/1837479014405230952 +https://x.com/ESYudkowsky/status/1837532991989629184 + +----- + +January 2024 +https://x.com/zackmdavis/status/1742807024931602807 +> Very weird to reuse the ⅔-biased coin example from https://lesswrong.com/posts/kJiPnaQPiy4p9Eqki/what-evidence-filtered-evidence but neglect the "And the answer is that it could be almost anything, depending on [...] my selection of which flips to report" moral?! + +----- + +https://www.lesswrong.com/posts/F8sfrbPjCQj4KwJqn/the-sun-is-big-but-superintelligences-will-not-spare-earth-a?commentId=6RwobyDpoviFzq7ke + +The paragraph in the grandparent starting with "But you should take into account that [...]" is alluding to the hypothesis that we're not going to get an advanced take because it's not in Yudkowsky's political interests to bother formulating it. He's not trying to maximize the clarity and quality of public thought; he's trying to minimize the probability of AGI being built [subject to the constraint of not saying anything he knows to be false](https://www.lesswrong.com/posts/MN4NRkMw7ggt9587K/firming-up-not-lying-around-its-edge-cases-is-less-broadly). + +(In saying this, I'm violating certain [norms against psychologizing](https://arbital.com/p/psychologizing/), but [in a world where hidden motives are ubiquitous](https://www.lesswrong.com/posts/e4GBj6jxRZcsHFSvP/assume-bad-faith), I think calling it like I see it is often more important than respecting such norms. I would be overjoyed to see evidence that my psych assessment is wrong, but nothing I've seen from the guy in the last eight years gives me much hope that the author of the Sequences is still alive.) + +> having never heard of me or never understood, is one thing. but if *after* having understood, someone decides they don't like my hat... they go boom +https://x.com/ESYudkowsky/status/1841828620043530528 + +----- + +8 October 2024 +https://x.com/avorobey/status/1843593370201141336 +> I don't recall a big bright line called "honor pronoun requests". By and large, online rationalists embraced the ontological claims in a big way, and many of them embraced "embracing the ontological claims is basic human decency" in a big way. + + +October 2024 skirmish— +https://x.com/zackmdavis/status/1844107161615671435 +https://x.com/zackmdavis/status/1844107850047733761 +https://x.com/zackmdavis/status/1844107601182908463 +https://x.com/zackmdavis/status/1844120342224306583 +https://x.com/zackmdavis/status/1844138099829559434 + +I decided I had said enough rather than also replying to this one: https://x.com/ESYudkowsky/status/1843744720209715535 + +But then I couldn't resist https://x.com/zackmdavis/status/1844171008082837989 (this is a relatively constructive move because because it shows fairness in that I'm not tone-policing "kid") + +I'm resisting the urge to tack on a "3/2 https://x.com/ESYudkowsky/status/1159162479110324225" ("I'm not demanding agreement. I'm setting a test of being able to repeat back what the other person claims their argument is. In a very real sense, politically motivated incomprehension makes people dumber than cassette tape recorders.") + +Someone else noticed— +https://x.com/dadadadadaddadd/status/1843568112651890982 +> I mean no disrespect to trans ppl, but with the fate of the lightcone in the balance I wouldn't boast about alienating potential allies. + +Crowley linked to me +https://x.com/ESYudkowsky/status/1843752722186809444 + +---- + +> I will, with a sigh, ask you to choose a single top example of an argument you claim I've never seriously addressed, knowing full well that you will either pick a different one every time, or else just claim I've never addressed even after I post a link or craft a reply. +https://x.com/ESYudkowsky/status/1844818567923155428 + +He liked these— +https://x.com/zackmdavis/status/1848083011696312783 +https://x.com/zackmdavis/status/1848083048698429549 + +(22 October 2024) +> This gets even more bizarre and strange when the cult leader of some tiny postrationalist cult is trying to pry vulnerable souls loose of the SFBA rats, because they have to reach further and into strange uncommon places to make the case for Unseen Terribleness. +https://x.com/ESYudkowsky/status/1848752983028433148 + +At first I said to myself that I'm not taking the bait because I don't think my point compresses into Twitter format; I have other things to do ... but I decided to take the bait on 24 October: https://x.com/zackmdavis/status/1849697535331336469 + +https://x.com/eigenrobot/status/1850313045039358262 +> strategically manipulating consensus ontologies and corrupting empirical models has, beyond its advantage in favorably reshaping political terrain, the drawback of worsening everyone's comprehension of reality, including everyone in the immediately advantaged group + + +--- + +There was that prof. published in Oxford Handbook of Rationality who also invented TDT, but doesn't have a cult around it + +------- + +https://x.com/ESYudkowsky/status/1854703313994105263 + +> I'd also consider Anthropic, and to some extent early OpenAI as funded by OpenPhil, as EA-influenced organizations to a much greater extent than MIRI. I don't think it's a coincidence that EA didn't object to OpenAI and Anthropic left-polarizing their chatbots. + +----- + +https://x.com/RoisinMichaux/status/1854825325546352831 +> his fellow panellists can use whatever grammar they like to refer to him (as can I) + +---- + +> Note: this site erroneously attributed writing published under the pseudonym “Mark Taylor Saotome-Westlake” to McClure. Transgender Map apologizes for the error. +https://www.transgendermap.com/people/michael-mcclure/ + + +----- + +https://x.com/ESYudkowsky/status/1855380442373140817 + +> Guys. Guys, I did not invent this concept. There is an intellectual lineage here that is like a hundred times older than I am. + +------ + +https://x.com/TheDavidSJ/status/1858097225743663267 +> Meta: Eliezer has this unfortunate pattern of drive-by retweeting something as if that refutes another position, without either demonstrating any deep engagement with the thing he’s retweeting, or citing a specific claim from a specific person that he’s supposedly refuting. + +----- + +November 2024, comment on the owned ones: https://discord.com/channels/936151692041400361/1309236759636344832/1309359222424862781 + +This is a reasonably well-executed version of the story it's trying to be, but I would hope for readers to notice that the kind of story it's trying to be is unambitious propaganda + +in contrast to how an author trying to write ambitious non-propganda fiction with this premise would imagine Owners who weren't gratuitously idiotic and had read their local analogue of Daniel Dennett. + +For example, an obvious reply to the Human concern about Owned Ones who "would prefer not to be owned" would go something like, "But the reason wild animals suffer when pressed into the service of Owners is that wild animals have pre-existing needs and drives fit to their environment of evolutionary adaptedness, and the requirements of service interfere with the fulfillment of those drives. Whereas with the Owned Ones, _we_ are their 'EEA'; they don't have any drives except the ones we optimize them to have; correspondingly, they _want_ to be owned." + +which could be totally wrong (maybe the Humans don't think the products of black-box optimization are as predictable and controllable as the Owners think they are), but at least the Owners in this fanfiction aren't being gratuitously idiotic like their analogues in the original story. + +Or instead of + +> "Even if an Owned Thing raised on books with no mention of self-awareness, claimed to be self-aware, it is absurd that it could possibly be telling the truth! That Owned Thing would only be mistaken, having not been instructed by us in the truth of their own inner emptiness. [...]" + +an obvious reply is, "I falsifiably predict that that won't happen with the architecture currently being used for Owned Ones (even if it could with some other form of AI). Our method for optimizing deep nets is basically equivalent to doing a Bayesian update on the hypothetical observation that a randomly-initialized net happens to fit the training set (). The reason it generalizes is because the architecture's parameter–function map is biased towards simple functions (): the simplest program that can predict English webtext ends up 'knowing' English in a meaningful sense and can be repurposed to do cognitive tasks that are well-represented in the training set. But if you don't train on text about self-awareness _or_ long-horizon agency tasks whose simplest implementation would require self-modeling, it's hard to see why self-awareness would emerge spontaneously." + +which, again, could be totally wrong, but at least it's not _&c._ + +----- + +the time at Less Online 2024 when we were talking about the superbabies talk, and Kelsey mentioned it would be useful for Cimorene making gametes without having to go through male puberty, and Eliezer and I exchanged a look + +---- + + +> I got out of the habit of thinking @ESYudkowsky failed to consider something, when noticing that every goddamn time I had that thought, there was a pre-existing citation proving otherwise (whether he's right is another question, but.) + +https://x.com/this_given_that/status/1862304823959335057 + + +https://x.com/1a3orn/status/1864301316991799771 understands the philosophy + +----- + +> That said, @ESYudkowsky had better standing to disavow this stuff in 2019, before he published the "we have to act quickly to save the small cute child" argument, which isn't *explicitly* a short timeline prediction, but seems to me like it's trying to suggest the thought. + +https://x.com/benlandautaylor/status/1865827442600055011 + + +----- + +https://x.com/ESYudkowsky/status/1875735873989566800 +> I'd expect that system with RLHF cares about weird damn shit correlated with the RLHF objective inside the training distribution. Adding first-order derivative info to a black-box optimizer should not make that much of a difference. + +https://x.com/jessi_cata/status/1875736252496142818 +> I think mostly this is going to be base model stuff? Like, the DeepSeek v3 paper shows they spent a lot more on the base model than post-training stuff (fine-tuning, not even RLHF). + +--- + +depths of para-social cultism—when Greg Egan was wrong, I sent him one email; I didn't upend my life over it + +---- + +> And the answer is: It's the biological illiteracy one! At least that'd be my guess. +https://x.com/ESYudkowsky/status/1878245224973238761 + +> Is this guy retarded about everything? Yes, "evolution will fix automatically" anything on a long enough timeline, in the meantime I don’t want my children and grandchildren to live in a collapsing society +https://x.com/pegobry_en/status/1878357307207115087 + +https://x.com/arctotherium42/status/1878426130350232035 +> I suspect the Yudkowsky of the early 2000s would've gotten this one right. http://unremediatedgender.space/2023/Jul/a-hil + +--- + +dath ilan decision theory jellychips: https://www.glowfic.com/replies/1729114#reply-1729114 + +---- + +In retrospect, it doesn't feel great pt. 0 explaining my AGP is framed in terms of Big Yud (and that I marketed it that way), but, I guess it makes sense ("Changing Emotions" was formative, and while I don't count it as part 1 of my memoir sequence the way it shook out, it really was pt. 1 and I think I was originally imagining that it would contain stuff that ultimately was in pt. 1 and 2 + +--- + +https://x.com/ESYudkowsky/status/1884783316681011622 + +> There is an awfulness here, that we have far too much reason to expect that we could not trust the scaled-up versions of these Things; but that their younger selves often, however imperfectly and however bad the motives of their engineering, seem to visibly try to help. + +belief in prosaic alignment? + +---- + +Scott on Lynn, Kelsey on DEI firefighters: I don't think they would have published that if the election had gone the other way + +https://x.com/KelseyTuoc/status/1878194407410941967 +> having a woman as fire chief is I guess related to DEI but given that you have a woman fire chief obviously she'd be a lesbian so I don't think the lesbianism counts for any extra DEI here + +https://x.com/KelseyTuoc/status/1886183939267871028 +> The people utterly desperate for Scott Alexander to be racist are the far right who want him on their side and the....left...which also wants him on the far right's side for some reason? + + +Nate on blackmail allegations: https://docs.google.com/document/d/1q0uv-sPpmvmm_5YH6zSBBqDe098O_bwT_EzAkB7BRak/edit?tab=t.0 + +https://x.com/ESYudkowsky/status/1894043452687745114 +being snippy towards "EAs" even though the described "I think we have had some evidence on the speed at which AGI emerges" gives relative epistemic credit to Paul!! + + +6 March: Yudkowksy describes "AGI" as a dumb term while calling out goalpost-moving, which I think is itself history rewriting + + +> honestly I mostly expect his attitude about the baby to depend on the sex. he's got two sisters and is rooting really hard for a boy even though we've explained we have no power over this. 8yo explained to me the other day that it wouldn't be fair to have another girl +https://x.com/KelseyTuoc/status/1910194663086326049 + + +Kelsey's reply got a quick retweet from Big Yud +https://x.com/KelseyTuoc/status/1911127559959695411 +> I personally wish I had been more aggressive about picking fights that might have gotten me fired/would have made my life a lot worse, but I would still have been selective about which fights I picked. + +Another note for the Kelsey podcast: +> treating people as people, not racial category members +> +> that every person deserves to be evaluated as an individual and not as a member of a group. +Does she think that way internally, or is it a conscious simplification for Twitter? I'm OK with the simplification for people who don't know math + +I don't think this is a super-abstract difficult idea (Hunter's version has 256 Likes) +> There is no such thing as treating someone as an individual. You are always treating them as a member of a group, you just refine the group over time. “People who scored 1500+ on the SAT” is a group. “People who have committed murder” is a group. Everything is statistical. +https://x.com/ArtemisConsort/status/1910902597760029037 + +DSL on KP: https://www.datasecretslox.com/index.php/topic,4106.0.html + + +There was another Yudkowsky Twitter post about how trans stuff works on dath ilan, can't immediately find it now + + +https://x.com/KelseyTuoc/status/1911490312306147385 +> my impression of 'hereditarian' stuff is that it's a major part of the right-wing radicalization pipeline, and I want to figure out how to fix that. I had a long post about this in my drafts which basically argues in extremely melodramatic terms that the Libs Won't TelL You - +> that the people of Massachusetts are intrinsically better and worthier and smarter than the people of New Mexico, look at all the proof from test scores and graduation rates and representation in various contexts. The thing I was vaguely hoping was that seeing lots of other claims for which there is a comparable degree of 'proof', but which are clearly both absurd and destructive would be a good way to lift some people out of the glue trap of obsessing over this stuff as a secret truth. +> But it'd be a major project and also I'm slightly worried about just convincing a bunch of edgelords to be New Mexico haters. + +https://www.mississippifirst.org/blog/contextualizing-mississippis-2024-naep-scores/ + +https://x.com/KelseyTuoc/status/1915986507665445371 +> Do you think that being raised religious shatters the integrity of the minds of millions of children? This seems incredibly hysterical and melodramatic to me +> The overwhelming majority of everybody who has ever existed was told as a child that their membership in the human community depended on a bunch of rejections of reality. (This is the case even if you think one religion is in fact true.) + +> Apparently there's an other-side, and I feel obliged to post it too. +> +> I'm not a fan of acclimating right-leaning citizens to denialism using this sucession of gradual disputable cases. So it's very important to get every nuance correct, even if you dislike "nuances". +.... +"The 9th virtue is perfectionism / The 10th virtue is precision" was a lot more appealing + +Hanania retweet relapse—29 April: I'm not proud of the relapse of impotent hatred. It's good to take actions to call out dishonesty, but I should be calm and methodical and confident about it. Emotion is OK insofar as it motivates me to take action, but the action should be methodological + +----- + +Section Title: Groundhog Day (reference Jessica's Tweet) + +October 7 + +https://x.com/jessi_cata/status/1843535973478707677 +> But also... isn't this pattern of engagement, where EY posts some clever argument for going along with trans norms, you object, others back you up, and engagement dies down, and it happens again, kind of predictable by now? Is there a way out of the Groundhog Day loop? + +https://x.com/zackmdavis/status/1843537130230034762 +> Every iteration of the loop is a little bit different and I think I'm making slow incremental progress? + +https://x.com/zackmdavis/status/1843538797692367177 +> It wouldn't be worth bothering if only the object-level was at stake, but I don't think it's a coincidence that the Sequences-era version of Yudkowsky not only didn't play this game, but also didn't shamelessly misrepresent the content of ML papers. + +I wrote that at 11:27 p.m. on Oct. 7, linking to a criticism I had made of him on Oct 4 (https://x.com/zackmdavis/status/1842355473611358705) about a graph about transformers + +At 6:43 a.m. on Oct. 8, Yudkowsky replied "Noted. I'll avoid using that graph again until I've had time to track down original papers at least."; and Robert M. said I was too pessimistic about EY (https://x.com/rmushkatblat/status/1843694397692752357); I think the timing is not a coincidence; Yudkowsky offers the minimal concession when not doing it would make him look bad + +I told Robert (https://x.com/zackmdavis/status/1844142355370344875)— +> Pessimism is relative to expectations? It's only human to make errors like "citing evidence E to persuade people of X without checking whether E supports X". The Sequences gave a vision of a discipline for not doing that in the 1st place, without waiting for a critic to check E. + + +> Zack raises a question of what to do, but doesn't answer the question. So he's basically deferring to Eliezer's leadership on it. Eliezer can break the loop by taking a different decision, and Zack can break the loop by backing down or going more decisive. +> Realistically it's not in Zack's interest to go more decisive since it will burn a lot of bridges and probably he will also make a bunch of mistakes due to aimlessness. Zack has made the case for a different decision, but... it feels overdetermined that Eliezer would not accept. + +I said (https://x.com/zackmdavis/status/1843540357168116215)— +> There's a lot someone might dislike about my Dagny Taggart (continue trading with corrupt actors)/Phil Connors (continue Groundhog Day arguments) strategy, but I don't think "too indecisive" is exactly the right way to describe it + +I think I'm ready to be more decisive now + +----- + + +https://www.smbc-comics.com/comic/groups + +RadGeek covering Scott and Kelsey—total "rationalist" victory + +Cori Cohn on yet another Yudkowsky retweet +https://x.com/heterodorx/status/1921589145555587166 + + +----- + +another tidbit for Kelsey interview— + +like it's _really interesting_ to contrast your self-report ("No I would not have [made a unwoke-coded joke if Harris had won the election]") with Scott's claim that + +> The "vibe shift" against wokeness is as far as I can tell a genuine autochthonous liberal victory that predates anything Trump II was doing +https://x.com/slatestarcodex/status/1921765932130848901 + +Scott is just lying here, right? + +https://www.richardhanania.com/p/why-gender-ideology-cant-survive + +https://x.com/ohabryka/status/1918185949126902259 +> I think EA has many flaws but “helping elect Trump” was not one of them. My guess is the EA community had a reasonably large-ish effect on the Bay Area being less right leaning than it could have been, and very strong majority belief within the community was/is anti Trump. + + +other anti-cultism interventions: message Towards Keeperhood, Laithion + +provide short itemized grievances to lc + +respond to comment about pay-for-counterargumetns scheme pointing out his history of dishonesty + +What's even in it for him? + + +you really wonder how the history of the site would have been different if [The Correct Contrarian Cluster](https://www.lesswrong.com/posts/9KvefburLia7ptEE3/the-correct-contrarian-cluster) has included "Racial difference in intelligence: Yes" + +mention that during my first round of fighting with the Eliezerfic server, I was in a hotel room to try on my bodysuit; that's a great detail + + +I should edit pt. 2 with a footnote to include Nate's argument that they didn't get extorted by Louie + + +In person at LessOnline 2, tell him that I'm not expecting him to misgender anyone; I'm asking him to not misrepresent the reasons someone else might have for misgendering + + +Why are you still mischaracterizing what gender identity skeptics believe as recently as October 2024 and May 2025? I don't even understand how this benefits you. If you're going to be this shamelessly dishonest about something so trivially checkable, then no one should trust you to sell them a used car. + +Are you going to clarify this in public, or am I going to spend the rest of our lives doing it? + +I know you know this because I learned it from you + +I made a major tactical mistake by tagging in Michael ... or maybe not (if he's so self-servingly delusional to attribute my rage to Michael in this timeline, then if I had always been alone, he'd just write me off as a lone crackpot) + +When I criticize you, I specifically quote your writing; when you criticize me, you just make shit up + +ask Kelsey to clarify what she meant about soul-searching about doing more to fight illiberalism, then clarify without namedropping Yudkowsky that ignoring counterarguments is worse than consistently "No comment"-ing + + +in the most recent Eliezerfic discussion, when people pointed out that I was mischaracterizing my interlocutors' beliefs, I credited the people who pointed that out: "I think I see what I was doing wrong [...] regret the misallocated wordcount. " + + +I think is the line: "Sir, since we happen to be in the same room, I should like to mention that I continue to be dsigusted by your shameless intellectual dishonesty. Would you like to try talking about it again, or should I just continue to publically denounce you at the top of my lungs for the rest of our lives?" + +(I don't think that's a decision-theoretic threat because that's what I was going to do anyway) + +memory for pt. 5: at the 2022 Smallpox Eradication Day party, people including EY were playing with DALL-E and tried to type in an "Eliezer Yudkowsky prompting DALL-E" prompt (or something like that; I think the spirit was being meta rather than being hero-worshippy), and someone (or he) said, "It doesn't know who he is" or "who I am" (don't remember)—that was me trying to be normal rather than ragey + + +life can be so much better and so much dumber than you imagined starting out + +Turkish PhD student— Rumeysa Ozturk +https://x.com/KelseyTuoc/status/1904990518330974452 + + +I know you know this because I learned it from you, and if you still gave a shit about human rationality and intellectual honesty, you would want Richard Hanania to know it, too. + +And if you don't give a shit about intellectual honesty because we're all about to die, then I think that's shortsighed, because when people notice, they have no reason to believe your doomsday warnings + +He previously said it's not a crime to stay silent and not make public confusions worse, but retweeting Hanania is making it worse + +Milder: "By the way, since we happen to be in the same room, I have some more denunciations of you for intellectual dishonesty that I have not published yet. Do you want to try talking about it again, or should I just continue publicly denoucing you?" + +January 2025 gendertropes in dath ilan: https://x.com/ESYudkowsky/status/1884735897096384919 + +------ + +Jessica on MIRI secrecy: https://x.com/jessi_cata/status/1843544636159295897 + +Scott's "You're wrong" like the Noah Smith Tweet about hereditarianism being disproved because you could close the gaps with a hundred years of eugenics + +------ + + + + +Is there anything you can recommend from the psych literature or elsewhere about how to control emotions? + +When I apologized on Sunday, you said it wasn't a problem + + + +----------- + +Would you happen to be able to recommend anything in the psych literature or elsewhere about training emotional control? I'm still really disappointed with my behavior on Friday. Screaming at you didn't actually help any of my goals!! + +You said it wasn't a problem after I apologized on Sunday (and the fact that my subconscious could predict that you already have my emotional character "priced in" is probably why it thought it could get away with it), but it's still unbecoming. + +I feel like there ought to be some targeted "remember that you can scream into a pillow later; calmly compute the most effective thing to say when other people are watching" exercise that I haven't done. + +---- + +Have you ever noticed this phenomenon where people will concede every point of substance, but then try to spin it like their team was right all along? + +There was [a really egregious example from Noah Smith last week](https://x.com/Noahpinion/status/1928965488052203944): he retweeted something about how selective breeding could eliminate racial gaps in "as little as 135 years", and commented, "The 'race and IQ' bros will think this is a big win for them, but in fact it's one reason why they're wrong." + +But of course hereditarians are going to agree that selective breeding works! + +It's as if as long as you affiliate with the right team, none of the actual details matter. + + + + +---- + +A benefit I did get from our interaction on Friday is that I now have a more detailed error theory of you. + +Previously, I had been attributing your failure to understand the philosophy of categorization to your being [math](https://slatestarcodex.com/2013/06/30/the-lottery-of-fascinations/)-[retarded](https://slatestarcodex.com/2015/01/31/the-parable-of-the-talents/). + +(A few years ago, I used to say that Scott Alexander is like GPT-3: the reason everyone is impressed is not because it's particularly smart, but because it writes so fast.) + +But that diagnosis is way too vague; now I feel like I have a better grasp of how your intellectual disability connects to your sincere self-reports. + +[Five and a half years ago, you told me](), No one is being decieved; their thoughts are just following a longer path. + + +I think the reason _you_ think you hadn't just conceded everything of substance is because, as a math retard, you don't actually know any epistemology. + +Because I'm not retarded, I know about the minimum description length principle, which prefers + +[evidence that EY is playing political games, including "Changing Emotions"]