Aaron Bergman's tweets, February 2023
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in February 2023 and has not deleted: 149 posts, oldest first, times in UTC. Each heading links to the original post on X. When Aaron replies to or quotes someone, their tweet appears as a blockquote with their name; those words are theirs, and everything outside the blockquotes is Aaron's.
This page is one month of the complete archive of Aaron Bergman's tweets, which updates automatically every few hours. Previous month: January 2023. Next month: March 2023.
Wednesday, February 1, 2023
Reply to @EzraJNewman, 00:48 UTC · 2 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
I don't think you're being fair Ezra - personally avoiding ≠ ostracizing (unless you also advocate/do it conspicuously)
Type I and II errors both matter too. If it's low-cost (few false negatives) to have high standards that's great
Reply to @EzraJNewman, 00:56 UTC · 2 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
1) thats terrible, I’m sorry
2) there are dimensions besides “single threshold level” on which to act. Like prisons need not be so inhumane, forms of consequence can be unbundled (like rn going to jail often causes losing job, social sanction, etc), and
Thread, continued, 00:58 UTC · 2 likes
least in theory you can imagine that the process of being accused (of anything) wasn’t *itself* a de facto form of punishment
Reply to @goth600, 08:03 UTC · 2 likes
@goth (@goth600): What stops lab grown meat from being sentient?
Uh not having neurons or a brain, just like plants
Reply to @larsiusprime, 08:26 UTC · 2 likes
Lars Doucet (@larsiusprime): Thinking about how "bodybuilding hard" and "math hard" fit best together.
"Math hard": impossible until it's done once, then trivial to copy.
"Bodybuilding hard" open to anyone to do anytime, but most people don't wanna because it takes constant effort/discipline.
There’s a related distinction: “hard” can mean either or both of (1) requiring of rare talent/skill and (2) unpleasant (often bundled with ‘requiring of willpower’)
These are very different concepts but get confused bc they’re often correlated in practice
Tweet, 18:41 UTC · 2 likes
By default I’m gonna assume this is trivial for everyone and already exists in a million places but…
Start: all-numeric dataframe, n outcome variables on the left, possible regressirs and controls on the right
Wrote some functions in R that, in 2 commands (can get down to 1),
Thread, continued, 18:41 UTC · 3 likes
generates a table of the estimates, p values, etc of every implied possible set of OLS regressions
Seems like there’s a 5% chance this would be useful to people and I should make a package?
Demo: https://aaronbergman.neocities.org/Rfunctions1
Reply to @peterwildeford, 19:04 UTC · 4 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford):
I disagree with you* but don't think you have bad epistemics
*in this post bc I don't think the latter is just a fancy way of saying the former and can think of both types of counterexample
Reply to his own earlier tweet, 19:19 UTC · 4 likes
Aaron Bergman, February 1, 2023: generates a table of the estimates, p values, etc of every implied possible set of OLS regressions Seems like there’s a 5% chance this would be useful to…
ppl I vaguely think are into #rstats / data science/econometrics (can tag whoever I missed) @frances__lorenz @StatsLime @Laura_k_Duffy @EzraJNewman @alyssamvance @inesferhumi @thatMikeBishop @ryancbriggs @redouad
Tweet, 23:48 UTC · 93 likes
A weird intellectual part of young childhood is that you have very little idea how smart adults are, and therefore how much smarter you’re going to be
Thread, continued, 23:48 UTC · 20 likes
It’s like if you knew you were growing a few inches a year but had knew only that adults were taller than you and not infinitely tall; mybe you'd get 10% taller, maybe 100%, maybe 100x – have to wait to find out!
Reply to @CharlesD353, 23:49 UTC · 2 likes
Charles🔸 (@CharlesD353): @EzraJNewman @StefanFSchubert @AaronBergman18 I think this ignores that an accusation is (maybe very weak, maybe quite strong, depends on context) Bayesian evidence and I think one should update accordingly depending on the circumstances.
Bayesianism doesn't tell you how to act, though. It's at least plausible that you should update but behave as though you didn't (for some size updates at least)
Thursday, February 2, 2023
Reply to @daniellellecco, 00:07 UTC · 2 likes
(The tweet being replied to by @daniellellecco is no longer available.)
lol, I mean at 12 or 14 sure but even at 6? 4?
Reply to @DefenseBased, 04:30 UTC · 2 likes
Everett Smith (@DefenseBased): If you don't like people, or you find them boring, try asking them better questions.
Most people, like large language models, can be surprising, helpful, or even delightful when given the right prompt.
*neurodivergent internet rationalists become 99th percentile charismatic upon realizing they can conceive of social interaction as prompt engineering*
Reply to @DKThomp, 04:40 UTC · 1 like
Derek Thompson (@DKThomp): This is the Super Bowl of all-time terrible Y axes.
Who ya got?
Problem on the right isn’t the y axis, it’s that their feet start at 4’9” and bodies about 1 foot tall according to it. That’s a graphics issue
Tweet, 05:45 UTC · 3 likes
NPR should make All Things Considered about cause prio
Reply to @panickssery, 08:17 UTC · 3 likes
Arjun Panickssery (@panickssery): Consequentialism is uncool (ie morally wrong) when framed as nerds committed to ethics that's always argued over empirically, but cool when framed as hardboiled "do what needs to be done" vs a weakness of will, a rule-following aim at denying one's responsibility for what happens
I don’t get what u mean by “uncool (ie morally wrong)” - but maybe that’s because I’m a nerd committed to empirically informed ethics
Reply to @AutismCapital, 08:29 UTC · 4 likes
Autism Capital 🧩 (@AutismCapital): Notice a similarity in the logos? It was there all along.
Proof Autism Capital is run by the Wikimedia Foundation
Reply to @moskov, 18:22 UTC · 7 likes
(The tweet being replied to by @moskov is no longer available.)
lfg
Tweet, 21:38 UTC · 4 likes
It’s a potentially useful term/concept and one I’ve used myself, but IMO “P(doom)” is not a good shorthand for “X-risky AI alignment failure”
Why? Bc in relative moral terms “doom” could be extremely bad (s-risk), very bad (prevents an amazing future), slightly bad,
Thread, continued, 21:38 UTC · 4 likes
(prevents an ok future), or good (replaces a bad future with a less bad or neutral one), scenarios which are generally—though not always—kinda insane to all group as one
Friday, February 3, 2023
Quote tweet, 01:26 UTC · 5 likes
I feel this - I’m pro-choice (under ~everyone’s understanding of what that term means) but my God the arguments (and “arguments”) that get pulled out in favor of ‘my side’ are terrible—even at the (nominally) Catholic university with an active+visible pro-life club I attended
Quoting Amos Wollen 🦐🪰✴️ (@goingawoll): Just got out of an Oxford Union debate on whether same-sex marriage should be allowed in the church. In a word - ABYSMAL. A thread:
Reply, 01:36 UTC · 1 like
I mean many people would probably wake up if a person walked by their bed. Seems like most sleeping arrangements make this uncommon
Reply to @reconfigurthing, 01:37 UTC · 19 likes
Elias Schmied (@reconfigurthing): i called again and the lady was annoyed that I called late and said it was fine and I should just let her roam😭i guess it's not like with dogs
also maybe oxford is small.. still tho she probably should've made it explicit on the thing that it's not that big of a deal
Maybe remove the missing sign then, lady?
Reply to @ruth_hook_, 04:13 UTC
(The tweet being replied to by @ruth_hook is no longer available.)_
I swear I’ve thought about making a very similar tweet. “Butlerian Jihad” would be such a good pejorative or ironic pro-woke co-opting of the pejorative
Thread, continued, 04:15 UTC
Wait what actually is it supposed to be? I haven’t read Dune either
Tweet, 04:47 UTC · 2 likes
FYI you can usually customize page size and set the length to like 10,000 inches so you never get page breaks
Reply to @absurdlymax, 04:52 UTC · 3 likes
Max Alexander (@absurdlymax): @AaronBergman18 Page breaks are good actually
Blocked and reported
Reply to @AndyMasley, 21:22 UTC · 2 likes
Andy Masley (@AndyMasley): Dune takes place centuries after an uprising against AIs that run the galactic society (the uprising is called "The Butlerian Jihad") and since the uprising it's "forbidden to make a machine in the likeness of a human mind"
🙏
Saturday, February 4, 2023
Reply to @DRMacIver, 09:41 UTC · 2 likes
David R. MacIver (@DRMacIver): Suppose you were given a completely empty white room to use as a bedroom. How would you go about deciding how to decorate / furnish / etc it?
Not just high level concepts, walk me through the steps you'd use. Where would you start?
Buy a mattress on Craigslist
Quote tweet, 10:47 UTC · 6 likes
Aw not again
Quoting Aaron Bergman 🔍 (@AaronBergman18): Time for the 3rd or 4th iteration of “Twitter rationalist forecasting nerds seem sufficiently concerned about something, should prob text my family”
Thread, continued, 10:47 UTC · 1 like
…
Reply to @peterwildeford, 11:26 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): It's possible for social media to be an immense positive instead of an immense negative. If you don't find it to be this way, you can make some changes! Or maybe it's just not for you!
Recently discovered that even Facebook - Facebook! - has crossed the neutrality waterline thanks to a growing proportion of EA/rat friends who evidently like to write
Reply to @RuxandraTeslo, 11:27 UTC · 2 likes
Ruxandra Teslo 🧬 (@RuxandraTeslo): It seems to me like Reddit is cringier than Twitter.
Ofc I’m biased but leaving that aside, and if you agree with me, what could be an explanation for it?
Cringe is often good tho. It’s (functionally) a sacrifice of reputation/status for the greater good
Quote tweet, 18:55 UTC · 8 likes
This is the correct take. Is EA a cult? Is chopped spinach a “processed food”?
Quoting Dan Elton (@moreisdifferent): Discussions about how EA has some cult-like properties raise some valid concerns, but they don't sway me much against EA movement.
A lot of things are cult-like - corporations, religions, political movements. It's more important to look what they are doing and what they believe.
Reply to @StefanFSchubert, 19:04 UTC · 3 likes
Stefan Schubert (@StefanFSchubert): I think people often overestimate how much political opinions are motivated by money.
E.g. I think how much NIMBY-ism is motivated by wanting to increase house prices is overstated.
How much anti-immigration views are due to fears of job losses is likely also overstated.
->
If my Econ of Immigration course was representative of that literature (and it was trying to be), economists uncritically subscribed to the second example too. Entire papers, popular models, and textbook chapters explaining immigration attitudes this way - zero consideration of
Thread, continued, 19:06 UTC · 2 likes
whether it makes any theoretical sense for a person to shape their views this way (since they have ~0 chance of affecting policy)
Tweet, 19:26 UTC · 5 likes
Idk who needs to hear this but if your (generic) prescription mysteriously isn’t getting filled, you might just have to call the pharmacy and tell them you’d rather pay out of pocket than wait for the outcome of a paperwork battle that would save you a grand total of $18
Thread, continued, 19:26 UTC · 3 likes
Once^ after inquiring I was told straight up they couldn’t fill it and had to ask about the option to pay Felt like a giga-agentic Live Player 😤😤😤
Sunday, February 5, 2023
Reply to @JesseTayRiver, 01:29 UTC · 2 likes
Jesse Smith (@JesseTayRiver): Most use of the word 'epistemic' is gratuitous. Very few instances where mental deletion of the term reduces the quality of writing. Let's just stop.
Most use of the word '' is gratuitous. Very few instances where mental deletion of the term reduces the quality of writing. Let's just stop.
Reply to @StefanFSchubert, 03:30 UTC · 2 likes
Stefan Schubert (@StefanFSchubert): People sometimes say one shouldn't use the word "should", but it seems to me a perfectly good word.
There are many things we should do, and which we should tell others to do.
Duties and obligations, and related emotions like guilt and shame, contribute to a better society.
My therapist a while back did not realize what she was getting herself into when she casually suggested I abandon normativity
Reply to @xsphi, 03:34 UTC · 2 likes
tess (@xsphi): am I the only one who finds turbulence really calming and pleasant?
Love that shit
Reply to @rickyflows, 15:34 UTC · 2 likes
ricky (@rickyflows): @personofswag I assume they’re a National Merit Semifinalist
I feel emotionally violated
Reply to @personofswag, 15:34 UTC · 2 likes
adam (6’4) 🇺🇸 (@personofswag): When someone says “the data are” I make a mental note that they must be a Very Smart Person.
What about “the datum is”
Reply to @noampomsky, 15:51 UTC · 2 likes
Ava (@noampomsky): when people talk in therapyspeak/like a non-violent communication textbook it signals a lack of emotional intelligence imo
I like my communication deadly
Reply, 16:43 UTC · 8 likes
Russia? No white folks there
Quote tweet, 18:09 UTC · 4 likes
Interesting thread but I disagree (short🧵)
Quoting kasra (@kasratweets): kinda interesting how for many years I just didn't get the point of reading fiction. I was like "why read that? it's made up, I want the TRUTH"
Quote tweet, 18:09 UTC · 2 likes
That is a naive view of truth indeed, but “a bunch of physics equations” is a naive view of nonfiction
Fictionality is correlated with but distinct from narrative/story-ness and entertainment/aesthetic value
Quoting kasra (@kasratweets): it's fine to want the truth...but there's something missing about this view. it's based on a naive view of "truth", as if truth is right there, in a bunch of physics equations, and isn't anywhere to be found inside a fictional story
Tweet, 18:09 UTC
Here’s “a bunch of physics equations” that’s also fiction:
1) E=.2mc³
2) g=4m/s²
3) K=1 Joule (always)
Thread, continued, 18:09 UTC
And here are a bunch of stories I’ve read+enjoyed that are also nonfiction
(Evidently I like the ‘financial crime narrative nonfiction’ microgenre but more below)
Thread, continued, 18:09 UTC
Book -> film -> both
Heads in Beds might be the single book I’ve most enjoyed reading
Thread, continued, 18:09 UTC · 2 likes
Also, as a matter of personal taste, I find “knowing the events actually took place” makes stories way more compelling
And while I don’t *mind* if anyone reads fiction, at least in a descriptive sense I think Yud was right that it’s not great epistemically (see disclaimer below)
Monday, February 6, 2023
Tweet, 04:27 UTC · 5 likes
First Manifold Market is live! (not sure why it took me so long) https://manifold.markets/AaronBergman/what-donationaccepting-entity-eg-ch?r=QWFyb25CZXJnbWFu
Quote tweet, 06:54 UTC · 3 likes
Initially pattern matches as bullshit to me but, nope, it’s spot on (well, I’m only confident about half, I’ll let you guess which) and extremely important for anyone who uses either of these (or similar) regularly to realize/plan/account for
Quoting dom (@ryiacy): Adderall makes people have higher agency, but think at low level of abstraction. Common pitfalls: being too risk-loving or getting bogged down in details
Weed makes people have lower agency, and think at higher level of abstraction. Pitfalls: being lazy, finding cliches profound
Reply to @cxgonzalez, 08:40 UTC · 2 likes
christian (@cxgonzalez): i pretty much never think about status idk why it’s such a concern for so many
I mean the main level of explanation is just evo psych. It’s a concern because the unconcerned had fewer kids
Reply to @mealreplacer, 12:32 UTC · 4 likes
(The tweet being replied to by @mealreplacer is no longer available.)
Sorry I won’t like it next time
Reply to @am_reports, 22:04 UTC · 2 likes
Ashley Miller (@am_reports): On this day in 1992: George Bush Sr sends out an all staff memo about his very very very very fat dog.
@natalia__coelho irrefutable proof of environmental contaminants theory
Reply to @NunoSempere, 22:08 UTC · 9 likes
Nuño Sempere (@NunoSempere): In my 20s, and I find myself making decisions like:
Where to live? Value of information dominates
Pick up friends/lover at the airport? Value of information dominates
New programming language? VOI dominates
Thinking of applying to a new job? VOI dominates
“Hey babe can u pick me up after my flight?”
“I’m sorry, I’ve never *not* picked you up after a flight, so need to see what will occur to help maximize the EV of all my future decisions”
Tuesday, February 7, 2023
Reply to @AFraserUrq, 00:46 UTC · 3 likes
Alastair Fraser-Urquhart (@AFraserUrq): Hot take: way way waaayyyyy too much pressure to be vegan within EA.
A community norm of something often incredibly difficult to do, for minimal welfare benefit, in opposition to most of the epistemic claims we make to triage our actions to do the most good, is a Bad Idea.
I agree wrt veganism per se, but not wrt “any pressure for animal welfare affecting personal diet”
Object level: buying a factory farmed chicken is bad
Signaling level: as things currently stand,
Thread, continued, 00:48 UTC · 3 likes
buying factory farmed chicken (at least if other consumption habits indicate adequate income) is pretty strong (though not irrefutable) evidence against “genuine impartial moral concern about others” which seems pretty important
Reply to @Laura_k_Duffy, 04:37 UTC · 3 likes
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): If you consider yourself an EA, are you a vegan? Vegetarian?
Vegan for 4 years, unfortunately don’t think it was (physiologically) sustainable for me
Reply to @gnrtvty, 04:43 UTC · 6 likes
kaiwan (@gnrtvty): @ollyrobot Looks like mostly no.
Sam Altman (CEO, OpenAI): no spouse, no kids
Demis Hassabis (CEO, DeepMind): married, two sons!
Dario Amodei (CEO, Anthropic): no spouse, no kids
Daniela Amodei (Pres, Anthropic): married to Holden Karnofsky of GiveWell, no kids (yet?)
Wait the Amodim aren’t married?
Reply to @eade_bengard, 04:52 UTC · 3 likes
Eade (@eade_bengard): Do I have any Greek followers who can tell me how to make pita all crazy fluffy like the first image? Second is mine
I too use a paintbrush to apply olive oil to a wheat balloon
Wednesday, February 8, 2023
Reply to @metakuna, 02:58 UTC · 2 likes
metakuna (14/100 blog posts) (@metakuna): I've been reading a lot of UK government documents and they're all actually very good. Narrative violation
They have a whole red teaming handbook that’s somewhere on my computer
Reply to @nolemonnomelon, 04:05 UTC · 2 likes
stove (@nolemonnomelon): seed oil appreciation thread what's everyone's favorite seed oil
Strawberry
Reply to @ruth_hook_, 04:07 UTC · 2 likes
(The tweet being replied to by @ruth_hook is no longer available.)_
Y’all planning Halloween in February?
Tweet, 04:13 UTC · 6 likes
Me and who
Reply to @jwmares, 04:18 UTC · 1 like
Justin Mares (@jwmares): Along with the disturbingly low rates of testosterone, there also happens to be historically low rates of marriage, birth, and sex.
Coincidence?
So is the 2015 fertility rate is 200ng/dL or is 2015 avg testosterone about 500 kids per woman?
Reply to @thebestestpie, 18:17 UTC · 2 likes
(The tweet being replied to by @thebestestpie is no longer available.)
Cope fs but https://www.aaronbergman.net/p/most-ivy-smart-students-arent-at
Is “they didn’t get in for reasons other than academics/intelligence?” still bad though?
Reply to @AndyMasley, 21:13 UTC · 2 likes
Andy Masley (@AndyMasley): Finding this made my day
If a Reddit anon ever calls me the 🐐of anything I will simply lie down outside and wait for my soul to be absorbed by the universe because my life would be complete
Thursday, February 9, 2023
Reply, 06:12 UTC · 2 likes
No
Quote tweet, 09:22 UTC · 5 likes
Learning how to stand up to bullies like me
Quoting Kareem Carr, Ph.D. (@kareem_carr): chapgpt is such a wuss lol
Reply to @ftlsid, 10:28 UTC · 7 likes
(The tweet being replied to by @ftlsid is no longer available.)
90% agree but not sure “harmless” is exactly the right word except in some very figurative sense
A bit more cynically, I think the (unconscious, functional) target is often some giga-Hansonian “signal X while also signaling that
Thread, continued, 10:28 UTC · 1 like
you don’t care too much about x and aren’t trying to signal it”
Tweet, 15:16 UTC · 27 likes
Get on my level😤
Reply to @wight1984, 15:50 UTC · 3 likes
Jamie H (@wight1984): @AaronBergman18 Is 'moves/talks/things in particular ways' an EA in-group reference to something?
I don’t think so, unless everyone else is in on something lol
I don’t really get the moves thing tbh, talks is largely just using relatively uncommon words and phrases more frequently than is typical (“plausibly”) but also being direct, other less legible things too…
Thread, continued, 15:55 UTC · 3 likes
Thinks is a bunch of things, but a few are
- high decoupling
- ontology (lol) incl. agents with goals, intrinsic and instrumental values, models, probabilities
- bayesianism
- probably just *more* intentional thinking
- using math/stats/econ/Phil/cs as lenses
Reply to @parafactual, 15:56 UTC · 3 likes
You (@parafactual): the average ea is not free. most are not for sale
But most are also are non-institutionalized citizens of relatively free states and societies
Reply to @SpacedOutMatt, 15:57 UTC · 2 likes
Matt 🔸 (@SpacedOutMatt): @AaronBergman18 Georgetown is elite and you know it
Maybe to some, not at EA Global
Reply to @EzraJNewman, 16:40 UTC · 3 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
Fun fact before a bunch of shit forced me into an indefinite (but hopefully not infinite) pause, I climbed pretty competitively for most of a decade
Reply to @AlexanderdeVri9, 19:33 UTC · 2 likes
Xander (@AlexanderdeVri9): Weak
I have like at least 19/25
Working on it
Friday, February 10, 2023
Reply to @peterwildeford, 07:20 UTC · 1 like
Peter Wildeford🇺🇸🚀 (@peterwildeford): You think space is cool, don't lie
I'm a centrist on the matter
Pros: its cool, no gravity, this website https://worldspaceflight.com/
cons: no air, cold, radiation, 3% fatality rate (source: https://www.worldspaceflight.com/ ofc), chatGPT won't make a picante take on the matter
Reply to @bryan_caplan, 18:55 UTC · 8 likes
Bryan Caplan (@bryan_caplan): Asking people to stop seeing family and friends to reduce their Covid risk?
Standard.
Asking people to lose weight to reduce their Covid risk?
Almost non-existent!
Per microcovid reduced, the former (I) is way easier for most people (2) has larger positive externalities and (3) (Im assuming but pretty confident) requires less messaging (per -mcv) likely bc “lose weight if you’re overweight” is super overdetermined
Reply to @peterwildeford, 18:59 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford):
Quote tweet, 19:54 UTC · 13 likes
Comments on this don’t seem very good but it’s a rare interesting and important question
Is there a rat-blogosphere canned answer or 8000 word ACX post I missed on this? If not, what’s your best guess answer?
Quoting Mehdi Yacoubi (@Mehdiyac): What do you think is causing this?
Reply to @Duderichy, 20:00 UTC · 8 likes
the Rich (@Duderichy): why is everyone dunking on this
(quoting @hedlike_a_hole: Bully these people into oblivion please)
Doing anything that requires more coordination than playing checkers is a cult
Quote tweet, 20:27 UTC · 6 likes
Don’t usually like to dunk (sorry @meaning_enjoyer) but IMO this is a perfect example of a criticism (of some activity or phenomenon, let’s call it X) that engages with X as though it were fundamentally a symbol (eg 🇺🇸) or work of art rather than an set of actions…
Quoting nihilism disrespecter (@meaning_enjoyer): @Duderichy it captures the hubris, arrogance and obliviousness of a certain class and it confuses class and status markers with virtues
Thread, continued, 20:27 UTC · 2 likes
intended to cause some change through normal-ass, non-symbolic, probabilistic Rube Goldberg machine-like set of channels
I don’t mean to say that symbolism/aesthetics/signaling aren’t important, just they’re often imprinted on something more concrete, like the paint on a hammer
Thread, continued, 20:27 UTC · 3 likes
And sure, (to continue the analogy), a hammer could be so ugly or have on it an image so disturbing so as to warrant throwing away an otherwise ‘perfectly good’ device, but almost all of the net value of almost all hammers comes from how good they are at pushing nails into wood
Sunday, February 12, 2023
Reply to @mirandahzhang, 13:07 UTC · 2 likes
Miranda Zhang (@mirandahzhang): How do you ensure you take actions aligned with your values?
(Not necessarily me lol, but in general and it’s what I endorse) macro-optimization>> micro-optimization. Transfer effort to the former as much as possible
And doing other than you’d do by default requires slack: time, money, energy, whatever
Both easier said than done ofc
Reply, 19:57 UTC · 4 likes
Jesus
Reply to @QualyThe, 20:04 UTC · 3 likes
Qualy the lightbulb (@QualyThe):
They should add “what’s the least impressive thing you’ve never done?” Like never shot a basketball? Never baked a cake?
Reply to @frances__lorenz, 22:25 UTC · 3 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
I woulda just shown up for work in Boston with some Hazmat suits and copies of The Precipice, tf they gonna do
Monday, February 13, 2023
Tweet, 10:21 UTC · 4 likes
Brb gonna train GPT-5 in PowerPoint
Thread, continued, 10:25 UTC · 2 likes
https://www.andrew.cmu.edu/user/twildenh/PowerPointTM/Paper.pdf
Nice
Reply to @ChamberVoice, 12:58 UTC · 5 likes
Chamber UK (@ChamberVoice): Labour accuse government of "lavish spending" on hotels, hospitality and other costs using taxpayer-funded debit cards 💳
🖼️ Treasury spends £3,393 on 13 fine art photographs
🍷 £23,457 on alcohol for UK embassies
✂️ Dept of Health spend £59,155 on stationery in March 2021
Ok I’m American and shouldn’t have a take, but these all seem perfectly reasonable
Tuesday, February 14, 2023
Reply to @ChanaMessinger, 08:07 UTC · 2 likes
Chana (@ChanaMessinger): @gnrtvty Mmm, this one seems more like trying to lie
I don’t think “lie” is right, though it’s sometimes disingenuous: “fault ownership” isn’t a well-defined thing and is often ambiguous, so the leader of a project isn’t ‘lying’ per se if she declares herself responsible for everything entangled with her leadership
Reply to @StefanFSchubert, 22:52 UTC · 1 like
Stefan Schubert (@StefanFSchubert): “Reviewing for free for a for-profit company is kind of a scam...prefers to review for non-profit journals
Seems to miss the key point. I think both for-profits and non-profits should pay for reviews.
The problem isn't that someone is profiting.
Actually yeah it is:
University pays academic to do things, incl peer review, also pays for journal subscription; receives results of both
Nonprofit journals would mean this loop is basically closed; for-profit means that there’s someone extracting rent each exchange…
Wednesday, February 15, 2023
Reply to @diviacaroline, 11:44 UTC · 1 like
Divia Eden 🔍 (@diviacaroline): Are there any serious arguments against the “modern landfills are fine it’s the trash that ends up not in landfills that’s the problem” take?
I don’t think I’ve heard anything convincing on that point, but I wanted to double check sicne so many people seem to think landfills=bad
I’m totally fine-pilled, but landfills don’t make any negative externalities to *creating* a bit of trash disappear. When written explicitly this sounds obvious, but maybe a pro-landfill ~vibe~ would decrease the facts salience idk
Thursday, February 16, 2023
Tweet, 10:45 UTC · 2 likes
I feel like I should know the answer to this but…
Black box: pull a lever on the side and a screen lights up indicating the outcome, which you know will be either H or T (not both)
Thread, continued, 10:45 UTC · 3 likes
You have no idea how this outcome is determined, but have no reason to favor expect one out outcome more the other, so your subjective probability of H is 50%
You pull the lever, and the screen shows H.
What, now, is your estimate of P(H) [i.e. credence of H on the next pull]?
Reply to @Theo_Clifford, 12:07 UTC
Theo Clifford (@Theo_Clifford): You're going to have to make some assumptions here! Lots of different ways to operationalise the idea that you have no other information.
One plausible answer Laplace's Law of Succession, where your estimate of P(H) should now be 2/3.
Hmm seems fake and wrong
Thread, continued, 12:10 UTC
Even if the first clause in that sentence is true in some case, the second seems totally made up. At most you can say that clause 1 excludes P=1
Reply to @absurdlymax, 18:10 UTC · 2 likes
Max Alexander (@absurdlymax): Any good examples of relationships, such as >, =, <, that are not those three?
Step outside of the space of real numbered vectors or smth and there are a bazillion qualitative ones you can’t (or I can’t) formalize:
- The relationship between the color red and the color blue
- between a thing and it’s name
- between me today and me tomorrow
Reply to @mealreplacer, 18:20 UTC · 2 likes
(The tweet being replied to by @mealreplacer is no longer available.)
Can I interest you in the greatest nerd snipe of all time?
Thread, continued, 18:21 UTC · 2 likes
https://en.wikipedia.org/wiki/Sleeping_Beauty_problem
I’m leaning towards either 1/2 or “no solution” at this point
Reply to @SpacedOutMatt, 18:37 UTC · 1 like
Matt 🔸 (@SpacedOutMatt): Dreaming of a new life
I could totally reuse my cover letter line about how philosophy minor taught me to “communicate…with clarity and precision”
Reply to @mattyglesias, 22:26 UTC · 2 likes
Matthew Yglesias (@mattyglesias): If someone wants to write an essay making the case that license plate laws are bad and should be repealed, I will read it.
But if we think cars should have license plates then you can’t just let people drive around with fake or expired paper tags.
I am legitimately confused by how bad of a take this is. Like I’ve probably read or heard you opine on literally thousands of things, and this is like sampling from a distribution that looked something like N(5,2) and getting -38
Thread, continued, 22:35 UTC
At least for laws that don’t land people in prison, it’s the rule rather than the exception that 100% success at literal enforcement would be pretty obviously terrible. Fine everyone going 61 in a 55? Citation for every 19 yo college student who drinks?
Friday, February 17, 2023
Reply to @OrionJohnston, 00:03 UTC
David Johnston (@OrionJohnston): I think reducing ambiguity in speed limits would be a good thing
I agree if it was made clear to everyone that would be the new normal!
Saturday, February 18, 2023
Tweet, 02:24 UTC · 4 likes
p-values would (generally) be more interesting and useful if H_0 corresponded to smth like “minimal interesting effect size” rather than “zero effect”
Thread, continued, 02:24 UTC · 2 likes
Like *given no publication bias* (a false assumption) p=.1 for “ketamine works better than Prozac” would IMO be more impressive than p=.02 for “ketamine works better than a placebo”
Tweet, 07:43 UTC · 8 likes
It really does try its best to be a good timer
(video, plays on X)
Sunday, February 19, 2023
Reply to @JacquesThibs, 00:21 UTC · 7 likes
Jacques (@JacquesThibs): It will be released on Monday
(quoting @Bankless: The AI episode with @ESYudkowsky we’re releasing on Monday is going to make waves. Is AI the humanity ender? Our early release discord is buzzing. We hope someone in the AI community can tell us why Eliezer is wrong. The convo starts Monday.)
I'm sure Eliezer has his reasons for going on this particular show, but I hope one of them is to signal 80k that he'd be willing to go on
Reply to @cornoisseur, 20:20 UTC
(The tweet being replied to by @cornoisseur is no longer available.)
Data scientists do *not* make $200k (wish they did tho)
Reply to @EzraJNewman, 20:35 UTC · 3 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
Depends on the nature of the simulation!
An excel sheet can be a simulation of a hurricane, as can attempted life-sized replica of the original storm
Tuesday, February 21, 2023
Tweet, 14:25 UTC · 9 likes
What’s that nootropic that makes people love socializing and talking? Piracetam maybe?
Yeah if someone could ship a lifetime supply to Yud that would be great. He’s got some podcasts to do
Reply, 15:03 UTC · 4 likes
Yes! But uh in that case I retract my call to action
Tweet, 20:03 UTC · 11 likes
I’ll bite. Who tf is Mr. Beast?
Thread, continued, 20:04 UTC · 4 likes
This is like “polycule” all over again
Wednesday, February 22, 2023
Reply to @TetraspaceWest, 05:05 UTC · 6 likes
(The tweet being replied to by @TetraspaceWest is no longer available.)
Classic EA Twitter, always responding to low effort tweets with an object level, citation rich, detailed review of the literature
Reply to @inesferhumi, 18:28 UTC · 7 likes
Inés (@inesferhumi): It’s happened
A mis padres no le gusta el Open AI
Thread, continued, 18:31 UTC · 2 likes
*les
damned if I didn’t learn how to conjugate gustar in 7 years
Reply, 18:33 UTC · 2 likes
FWIW Yud doesn’t think other animals are sentient
Thursday, February 23, 2023
Reply to @absurdlymax, 13:31 UTC · 3 likes
Max Alexander (@absurdlymax): What are the best arguments that Toby Ord's 1/6 estimate of extinction this century is too high?
Probability is fake and the real answer is 0 we just can’t know it rn
Reply to @ESYudkowsky, 18:55 UTC · 16 likes
Eliezer Yudkowsky ⏹️ (@ESYudkowsky): Despite everything I know this still brought tears into my eyes.
Explanation buried elsewhere in comments https://x.com/the__unraveling/status/1628812845763776512?s=46&t=JAwYV59AMTBN5LgmRzUjIw
Reply to @BenHayum, 19:07 UTC · 7 likes
(The tweet being replied to by @BenHayum is no longer available.)
Him and Hanania going from worst person on the internet to weirdly ambiguous Severus Snape energy
Friday, February 24, 2023
Reply to @sama, 10:41 UTC · 12 likes
Sam Altman (@sama):
Vibes here are like when $6.5T in market cap just poofs into existence because Jerome Powell had lunch with a guy who’s PhD advisor called for lower interest rates in Canada in 2003
Thread, continued, 10:43 UTC · 2 likes
Not complaining tho
Reply to @EzraJNewman, 16:23 UTC · 2 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
1) go to amazon
2) look up “shorts”
3) select one of the options
4) proceed to checkout
Reply to @jam3scampbell, 16:25 UTC · 2 likes
(The tweet being replied to by @jam3scampbell is no longer available.)
I hereby second this request. Can’t say no now @sama
I don’t make the rules.
Reply to @EzraJNewman, 19:08 UTC · 2 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
Forget the OWID charts, the sheer variety of shrimp shorts I could (did???👀) buy while comfortably holding my breath is easily the best argument for capitalism
Reply to @SpacedOutMatt, 19:12 UTC · 1 like
Matt 🔸 (@SpacedOutMatt): I'll defend everything but R:
Matlab: genuinely unparalleled number of features for analyzing certain engineering problems. Incredible quality of documentation.
Fortran: A beautiful language from a more civilized time
C++: Try writing python code that can run on a satellite
Blocked and moral weight reduced according to cognitive capacities un-demonstrated
I don’t know if God exists, but if He does He surely speaks in R
Reply to @SpencrGreenberg, 19:55 UTC · 7 likes
Spencer Greenberg 🔍 (@SpencrGreenberg): There's a deep sense in which money can't be "wasted" from a bird's eye perspective - only resources and people's time can be wasted.
If someone "wastes" $100, someone else now has $100 extra to spend. Even burning bills deflates the currency, making other bills more valuable.🧵
99% agree, but I think this misses the *intrinsic* value of money as a social technology for facilitating trade. Easy to see in the extreme: burn *all* the money, and (I claim) this new world is less materially wealthy because people value liquidity and now there is none
Reply to @parafactual, 20:53 UTC · 7 likes
You (@parafactual): like a dozen people reviewed the draft
Quite the list
Reply to @mbateman, 21:15 UTC · 7 likes
Matt Bateman (@mbateman): We have a rule in our schools: there aren’t special kinds of moral work that merit moral gold stars.
If you organize a neighborhood cleanup or do volunteer work, great.
If you get a job at Baskin Robbins or play in a band or write an app… also great!
*Exactly* just as great.
Don’t get me wrong, most “moral work” done by teens that is ~100% signaling and learning (not bad things btw), and in these cases I agree. But do you really want to foreclose the possibility of giving praise/acclaim for things that are, like, actually very morally good?
Saturday, February 25, 2023
Tweet, 04:53 UTC · 12 likes
Are there recent historians? People who specialize in August, 2014 instead of the 18th century?
Like at some point “study of the recent past” becomes “journalism,” but it seems like between, say, 2 and 10 years ago is nobody’s job
Recent History Twitter lmk if I’m mistaken
Reply to @acheronix, 05:11 UTC · 2 likes
Acheronix (@acheronix): @AaronBergman18 https://knowyourmeme.com/
Doing Gods work
Reply to @CrispyWicks, 05:46 UTC · 2 likes
Chris P. Wicks 🔎🐜 (@CrispyWicks): Gone to extravagantly fancy sushi place, intimate omikase. Seated next to 14-year married couple. Our own relationship 6 months deep. Considering what 14 years would look like. Double that, double *that*. Lot to consider.
Grandparents made it to “*that*” and more than a decade beyond. Biology stopped them just shy of reaching 70 years married (not sure about in total). Nothing especially profound to say- just that long, happy relationships (of any kind, really) seem like a wonderful thing
Reply to @tyler_m_john, 09:22 UTC · 4 likes
Tyler John (@tyler_m_john): Managing to fully align a chatbot by training it on Tinder data so it just says "hey"
Beautifully symmetric informational asymmetry
Reply to @peterwildeford, 20:29 UTC · 3 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18 @SpacedOutMatt If God speaks R that would explain so much about why the world doesn't work. Now that you mention it, the design of the world shares a lot of similarities to the design of R in how barely patched together it is.
“Sir, it seems the agents in sim 21518876a have figured out their sim’s source code”
Reply to @GarrettPetersen, 21:31 UTC · 8 likes
Dr. Dad, PhD 🔄🔼◀️🔽▶️ (@GarrettPetersen): My hot dating take for today is that attractiveness is at least somewhat absolute. So if everyone got 10% more or less attractive, it would have meaningful effects beyond everyone just raising or lowering their standards proportionally.
I think this is mostly true, but some (even physical) dimensions of attractiveness I bet are subconsciously defined in relative terms. Most obviously height, since subpopulation averages differ so much - like attr(5’8”) is probably really attr(40th observed percentile)
Sunday, February 26, 2023
Reply to @Noahpinion, 01:24 UTC · 8 likes
Noah Smith 🐇🇺🇸🇺🇦🇹🇼 (@Noahpinion): Corbyn was always bad and I always told you he was bad, and I was right
(quoting @OzKaterji: Jeremy Corbyn’s post-Labour career continues to plumb new depths as he met with the racist conspiracy theorist Roger Waters last night.)
I don’t think “taking picture with bad person is sufficient for badness” is true or good to promote
Reply to @frances__lorenz, 19:35 UTC · 10 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
Part of an old Substack draft titled “How is everyone in EA so goddamn competent?”
Monday, February 27, 2023
Reply to @ohwizenedtortle, 23:57 UTC · 5 likes
oh dandy lion🔸 (@ohwizenedtortle): Absolutely bonkers that Peter invented the idea of rethinking things. It's such an obvious idea to not only ever believe one thing and never change your mind even when given evidence. I think lots of people, especially those who have to engage in the world, should try this.
(quoting @peterwildeford: Happy to announce our newest incubated project from @RethinkPriors ! (joke - product is real but not associated with us))
I hereby invent Rethink Rethink to brainstorm how RP could be better, and Unthink to scrap any bad programs at Rethink, and Unthink Rethink to return RP to its original mission, and Rethink Unthink to improve Rethink more efficiently, and Unthink Unthink once RP can’t be improved
Tuesday, February 28, 2023
Reply to @mealreplacer, 00:05 UTC · 4 likes
(The tweet being replied to by @mealreplacer is no longer available.)
somewhat bizarre that I can read the two extremely common words “Paul slow” , which doesn’t make any intrinsic sense and afaik I’ve never seen before, and know exactly what it’s referring to
Reply to @maiamindel, 18:52 UTC · 7 likes
Maia (@maiamindel): Why do gyms have so many checks at the door: IDs, signing in and out, making appointments, etc
Bro the people that *pay* for a membership basically never go, who's breaking in there?
Me. Used different email addresses to sign up for a one day free trial with a gym 6 or 7 times. My dumb ass used eg ab1@gmail, ab2@gmail, etc. Eventually the manager must have gotten suspicious and figured things out, because he found me about to start working out and accused…
Thread, continued, 18:53 UTC · 6 likes
me of doing just that. I was like “yup, you caught me. I’m gonna leave now” and got tf out of there in like 4.5 seconds
I do actually think this was somewhat unethical tho, kinda feel bad
Reply to @maiamindel, 19:07 UTC · 5 likes
Maia (@maiamindel): @SpacedOutMatt @AaronBergman18 17 octillion potential shrimp QALYs die for every dollar you spend
I’m a pure hedonic utilitarian except for also intrinsically caring about the financial well-being of fitness center owners in wealthy countries
End of February 2023. Previous month: January 2023. Next month: March 2023. Back to the archive index.