Aaron Bergman's tweets, March 2023
Every public tweet, reply, quote tweet and retweet that Aaron Bergman (@AaronBergman18 on X, formerly Twitter) posted in March 2023 and has not deleted: 212 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: February 2023. Next month: April 2023.
Wednesday, March 1, 2023
Reply to @mealreplacer, 01:58 UTC · 4 likes
(The tweet being replied to by @mealreplacer is no longer available.)
… “namely Ana here who’s been talking my ear off lately about how interpretability is the most promising research direction in AI safety and Liza who thinks it’s utterly futile and is instead bullish on the natural abstractions framework”
Reply to @woj_zaremba, 02:02 UTC · 3 likes
Wojciech Zaremba (@woj_zaremba): Civilization altering technologies tend to scare many people.
Consider electric current - an invisible force that can kill a human on the spot, demonstrated to kill an elephant.
Are you willing to have it embedded in your house walls, next to your children?
Reply to @acidshill, 04:08 UTC · 1 like
shill 🔍 (@acidshill): sometimes i think about what the web could have been, what it should have been
i don't expect this to ever happen, for a few reasons, but i think it's technically possible with e.g. ipfs
(quoting @acidshill: and i don't think we've really seen everything hypertext can do just yet)
Neocities might be a very simple example of what you’re gesturing at? At least, the idea seems to be “social media bounded only by the limits of html/css/js” http://neocities.org
Tweet, 05:18 UTC · 30 likes
The store brand olive oil people have short timelines and not much hope, it appears
Reply to @AndyMasley, 18:23 UTC · 2 likes
Andy Masley (@AndyMasley): @absurdlymax Risks making people think that probability is describing Everett branches, which it definitely isn't. Imo it's much better (but more clunky) to just say "It's most rational based on my limited information to assume this situation is like rolling a 10-sided die" or something.
Extremely inclined to defer to the physics teacher here, but supposing the Everettian interpretation (?) is correct, wouldn’t probability then indeed correspond to proportions of branches?
*Your* dice is >99.99% Newtonian/deterministic, but the information you have doesn’t…
Thread, continued, 18:24 UTC · 2 likes
tell you which set of branches you’re in prior to rolling: the set in which the dice comes up 1, in which it comes up 2, etc.
Reply to @AndyMasley, 18:38 UTC · 2 likes
Andy Masley (@AndyMasley): I might be wrong but my understanding of the Everett Interpretation is that in this case almost all the branches would go to one specific die roll? They're clustered around what would happen in a deterministic universe where everything's preset? I could be totally wrong here.
This is probably butchering the model but:
-dice roll at t=0
-A:= set of branches that exist at t=-1
-a:= your unique branch at t=0
I agree that all the branches that started from a will cluster basically deterministically, but I’m thinking that your limited info corresponds to
Thread, continued, 18:41 UTC · 2 likes
Which of the branches in A, say {a, b, c,…z} you could or could not be in based on your observations
Like you know you’re not in the one where the dinosaurs kept on evolving bc of some weird fluctuation in the early universe that altered the asteroids path…
Thread, continued, 18:42 UTC · 2 likes
But you don’t know whether you’re in the one where your hand is slightly more firm or slightly less firm because of some random quantum event somewhere, and *those* yield different deterministic outcomes
Quote tweet, 20:41 UTC · 41 likes
I am extremely anti-anti-random tiktok girl
1) Being mean is bad *even if the person you're being mean to uses filler words in a way you don't like*
2) I'd bet that "eloquence in front of a live mic" is almost unrelated to intelligence and even off-mic verbal clarity
Quoting wilder (@wilderpatriot): Like wtf is she talking about? 😂
Thursday, March 2, 2023
Reply to @jide_alaga, 01:36 UTC · 5 likes
(The tweet being replied to by @jide_alaga is no longer available.)
Equally likely (p=0) because I am neither Sam nor black
Tweet, 05:53 UTC · 12 likes
Freeze dried strawberry powder tastes ILLEGAL. Like this is *precisely* what evolution was optimizing for
My ancestors from the caves and the shtetl are offended rn
Reply to @epiplexity, 13:34 UTC
(The tweet being replied to by @epiplexity is no longer available.)
Eh this depends on an almost tautological definition of fairness like “that which emerges from a competitive marketplace of rational actors”
Regardless of how appropriate that definition is, I don’t think it’s the one most ordinary people implicitly use
Reply to @RuxandraTeslo, 13:41 UTC · 2 likes
Ruxandra Teslo 🧬 (@RuxandraTeslo): What fascinates me abt a lot of social science research is that results which end up being true/replicable are things one could have known simply by living in the world and meeting people (& maybe being cultured), whereas most of the surprising stuff turns out to be fake.
(quoting @JonHaidt: 5. Check out the evidence of causality for yourself. I organize more than a hundred published studies, on all sides, in this post on the After Babel Substack. Subscribe (for free) to get more: https://jonathanhaidt.substack.com/p/social-media-mental-illness-epidemic)
I think what seems obvious and surprising to college educated twitter users with firstnamelastnameDotCom in their bio is downstream of social science more than this implies. Endogeneity!
Reply to @zetalyrae, 14:38 UTC · 1 like
Fernando Borretti (@zetalyrae): One of the skills I admire most: the ability to take a seemingly complex problem and break it down into a simple yet robust model that uses little more than high school math, a couple probability distributions and some logarithms. You see this a lot in @anderssandberg papers e.g.
Just observing: this is the opposite of a bunch of LLM capabilities. Take addition:
“ simplify 14+51”
“ok lemme read 90% of the words ever written in the universe and, using thousands of dollars of electricity, 50 years of CS research, and 160 engineer IQ points,
Thread, continued, 14:40 UTC
search all of 730-dimensional vector space corresponding to an algorithm that will do something resembling ‘next word prediction’ which happens to sometimes correspond to correct arithmetic because simple math problems on the internet are usually correct”
Reply to @cafreiman, 14:44 UTC · 8 likes
Chris Freiman (@cafreiman): Elizabeth Warren defended raising taxes on the rich partly on the following grounds: “There is nobody in this country who got rich on his own . . . I want to be clear: you moved your goods to market on the roads the rest of us paid for.” 1/
Circular counter-argument, no?
1) Warren says taxes should be higher to repay social debt
2) you say social debt doesn’t exist because the rich already pay taxes
The whole question is about the right tax level
Reply to @zswitten, 14:50 UTC · 1 like
Zack Witten (@zswitten): Trying to find the tiniest image representations possible. I have a theory if it fits in 2K tokens Sydney will be able to visualize it.
You can write images straight into html like this https://stackoverflow.com/questions/47533358/base64-icons-favicon
Reply to his own earlier tweet, 18:09 UTC · 2 likes
Aaron Bergman, March 2, 2023: Circular counter-argument, no? 1) Warren says taxes should be higher to repay social debt 2) you say social debt doesn’t exist because the rich already pay…
Guys don’t get me wrong I don’t like Warren either and think she’s wrong on this lol
Friday, March 3, 2023
Reply to @peterwildeford, 03:25 UTC · 4 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): @ohwizenedtortle Unfortunately I only date human women, not turtles
Consider, per this Rethink Priorities report, whether this may be due to speciesism arising in part from a biased perception of turtle happiness and autonomy https://static1.squarespace.com/static/6035868111c9bd46c176042b/t/60bd143940b94d38ea45cc34/1623004217909/How%2Bto%2BMeasure%2BCapacity%2BReport%2B__%2BRethink%2BPriorities.pdf
Reply to @ohabryka, 04:20 UTC · 4 likes
Oliver Habryka (@ohabryka): I feel like they didn't really make any great arguments. There is a lot of writing on ACX and SSC and LessWrong on the unreliability of the press, and indeed journalists seem on average much more interested in telling a salacious story than bloggers.
I think some credit is due for a basic level of good faith engagement but yeah I agree. IA may be kind, smart, and earnest, but every criticism I’ve seen of his seems basically just wrong and not in a productive or interesting way
Tweet, 06:23 UTC · 2 likes
A ranked choice vote doth hereby called
🐐:
Thread, continued, 06:23 UTC · 1 like
🥈:
Thread, continued, 06:23 UTC · 1 like
I briefly considered starting the thread with “and now it’s time to rank women on the internet” but thought that would be a smidge too edgy, so I’m making the joke in a very roundabout way right here
Reply to @Laura_k_Duffy, 14:29 UTC · 9 likes
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): Announcing: EA Twitter Cookbook! 🎉🎉🎉🎉 (with @baesrule1 and @ryancbriggs)
If we get enough submissions, I’ll make an e-cook book for everyone!
What are your favorite recipes?
Link them/post pictures below!!To start: Kimchi fried rice, but with udon noodles instead 🍜 😊
1) acquire potato
2) microwave for 4+z_p minutes where z_p ≡ z score of potato in the distribution of potato masses
3) open microwave
4) reach inside
THREAD CONTINUES BELOW
Thread, continued, 14:36 UTC · 4 likes
5) rotate potato 180° in any two dimensional vector space spanned by <0,0,1> z axis and <a,b,0> where a and b are real numbers of which at most one is 0
6) close microwave
7) repeat step 2, with time reduced by 1 minute
8) allow potato to cool to reduce risk of injury or death 🔚
Reply to @SpacedOutMatt, 14:37 UTC · 2 likes
Matt 🔸 (@SpacedOutMatt): @AaronBergman18 @Laura_k_Duffy @baesrule1 @ryancbriggs Honestly tho baked potatoes are underrated
🥔
Quote tweet, 14:46 UTC · 11 likes
👍
Quoting @goth (@goth600): Gm
Reply to @frances__lorenz, 15:41 UTC · 2 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
100% unironically Microsoft Word is the best amateur graphic design program
Reply to @AlecStapp, 15:55 UTC · 5 likes
Alec Stapp (@AlecStapp): If it were good for their business, then why would the government need to mandate it 🤔
Not in this case but there are possible reasons:
- coordination/first mover problems
- needing someone to blame for pr reasons
- increased demand from perceived credibility of mandates
Reply to his own earlier tweet, 16:48 UTC · 4 likes
Aaron Bergman, March 3, 2023: 5) rotate potato 180° in any two dimensional vector space spanned by <0,0,1> z axis and <a,b,0> where a and b are real numbers of which at most one is 0 6)…
Update: didn’t follow step 8; [maybe a couple epidermal skin cells] died
Mugshot of the perp👇
Reply to @dwarkesh_sp, 22:05 UTC · 7 likes
Dwarkesh Patel (@dwarkesh_sp): What should I ask Scott Aaronson?
He is the world's leading theoretical quantum computing researcher.
He's recently been working with OpenAI on AI safety - where he came up with the GPT watermarking scheme.
He also happens to be my former professor.
In case you didn’t stumble across this (not sure if he’s discussed elsewhere) and want college admissions takes from someone who’s taught (I assume?) at both Ivys and non-Ivys https://www.aaronbergman.net/p/most-ivy-smart-students-arent-at/comments
Reply to @ohwizenedtortle, 22:09 UTC · 4 likes
oh dandy lion🔸 (@ohwizenedtortle): i find it good to poke holes in potato on both sides with a fork first! lets the steam out which makes it not splat/explode
That may be “standard agricultural practice” and therefore exempted from all welfare legislation, but IMO it’s only ethical to cut the potato *after* it’s been stunned by the radiation
Reply to @cafreiman, 22:45 UTC · 17 likes
Chris Freiman (@cafreiman): This is cool—a market for food banks
Hear me out: a market for food bank *recipients* but it’s just the normal market and we give poor people money
Saturday, March 4, 2023
Reply to @ruth_hook_, 02:39 UTC · 10 likes
(The tweet being replied to by @ruth_hook is no longer available.)_
Can u do this but for guys
Tweet, 03:01 UTC · 41 likes
I am calling for a total and complete shutdown of Taylor Swift, Hinge, and—just to be safe—heterosexuality until we can figure out what the hell is going on
Tweet, 03:01 UTC · 15 likes
Do all adult female humans express such behavior, is the algo just doing its thing, or is this the AGI’s way of setting me up with Taylor Swift?
Thread, continued, 03:07 UTC · 14 likes
All this btw is from like 2.5 hours over 3 days, and I’m 110% sure I missed a bunch.
Nothing else comes close - *maybe* Taylor≈reading+hiking+coffee
Reply to @ohwizenedtortle, 03:55 UTC · 2 likes
oh dandy lion🔸 (@ohwizenedtortle): @AaronBergman18 wait aaron do you not like taylor swift??
No I do like Taylor Swift! But evidently not enough for me to intuitively expect/understand why she shows up way more than any other noun lmao
Thread, continued, 03:57 UTC · 1 like
She’s even on my repeat playlist (basic but solid lineup btw, all recommended)
Reply to @paulg, 05:12 UTC · 4 likes
Paul Graham (@paulg): One of the things AI coding assistants are showing us is how much duplication there is in the code that's out there. Most people just write the same stuff over and over.
The world only really needs read_csv, mutate, group_by, summarize, and maybe left_join
Everything else (yes, in every language) is a linear combination of these
Reply to @EzraJNewman, 12:08 UTC · 3 likes
(The tweet being replied to by @EzraJNewman is no longer available.)
Not for long if my plan to memorize all Taylor Swift songs goes according to plan
Tweet, 12:22 UTC · 14 likes
10/10 no notes @MatthewAdelste1 https://benthams.substack.com/p/takeaways-from-ea-global
Reply to @frances__lorenz, 13:37 UTC · 3 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
Indeed, and yet so few likewise mention Miguel de Cervantes who popularized the narrative itself with his ur-novel Don Quixote, suggesting that something besides aesthetic quality and sheer cultural influence accounts for Swift’s frequent invocation
Reply, 20:52 UTC · 3 likes
Oh lmao, but I neither confirm nor deny
Reply to @catehall, 21:03 UTC · 12 likes
(The tweet being replied to by @catehall is no longer available.)
As a patient I’d honestly prefer 80th percentile MCAT and 20th education to the reverse
Sunday, March 5, 2023
Reply to @FrostedSapling, 00:01 UTC · 6 likes
Ivan 🔸 (@FrostedSapling): @AaronBergman18 @catehall You do realize that they take the MCAT before they get any medical training right? Might as well go by their SAT scores
Yeah, I would say the same substituting in SAT
Reply to @FrostedSapling, 00:06 UTC · 2 likes
Ivan 🔸 (@FrostedSapling): I'm not saying education is the be all end all anyway. Residencies problem matter far more, but in reality no person can really tell a good doctor from a bad one from the outset
Setting ER and surgery aside (which is a big category ik), I'd gladly take someone very smart with access to UpToDate and a library over an arbitrary IRL doctor. Maybe 80th percentile MCAT is too low idk. Definitely 99th percentile tho
Reply to @_chair, 00:12 UTC · 2 likes
Chair (@_chair): What's her @
@taylorswift13 but if you mean her snap I'm afraid I can't reveal that publicly
Reply to @natalia__coelho, 00:20 UTC · 2 likes
(The tweet being replied to by @natalia__coelho is no longer available.)
No and wouldn't literally experience the emotion of guilt, but would say I'm wrong for not doing it and ought to do that
Reply to @Laura_k_Duffy, 00:21 UTC · 2 likes
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): Also Caramel can’t wait to meet all you DC people!
That's not a camel thats a dog
Thread, continued, 00:22 UTC · 1 like
It appears I misread the tweet :(
Thread, continued, 00:23 UTC · 3 likes
but still endorse the truth of my claim
Reply to @ESYudkowsky, 00:25 UTC · 14 likes
Eliezer Yudkowsky ⏹️ (@ESYudkowsky): I should probably try another podcast. I suspect it'll be hard to duplicate that interaction, but "a hypothesis affords testing". But I don't know where to find a podcast where the host asks questions the host is actually interested in, rather than being like...
I am a simple man. I see a reply here mentioning @dwarkesh_sp; I hit like
Reply to @frances__lorenz, 00:51 UTC · 2 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
Three plausibly-true but unrelated propositions. I wonder what she is trying to say🤔
Reply to @spiantado, 01:33 UTC · 3 likes
steven t. piantadosi (@spiantado): Every "philosophical" problem is just people failing to recognize that words don't have precise definitions, and words get used in different ways.
Once we give words precise definitions, "philosophy" turns into just ordinary mathematics.
This is a super common failure mode, but def disagree that it’s (necessarily?) true for every problem
Some things rational people would keep debating the existence of even with perfect semantic clarity: qualia, causality, true normative propositions
Reply to @ZerothAxiom, 02:23 UTC · 3 likes
Colby (@ZerothAxiom): @metakuna Physical attack is an ancient phenomenon and historically a somewhat likely way to die
So I do assume that frequently experiencing a thing that influenced the evolutionary development of our emotions and motivations might help us regulate those emotions better
It does regulate them better - but better from the perspective of evolution in a dangerous environment looks like “be super anxious so you don’t die”
Reply to @dissproportion, 03:05 UTC · 2 likes
(The tweet being replied to by @dissproportion is no longer available.)
I hate myself for using it in the 90% of cases it just doesn’t sound right
Thread, continued, 03:11 UTC · 2 likes
🙂🔫
Tweet, 06:02 UTC · 12 likes
It seems the avant-garde of twitter rationalist nerds has turned against the paper I RT’d being novel and important, so I deleted this tweet
Luckily I can piggyback off of them so never have to do any object level research whatsoever
Reply to @relic_radiation, 17:42 UTC · 4 likes
Elena Lake 🌿 (@relic_radiation): massage tip:
barring individual quirks and preferences, basically any touch that is both confident and attuned feels good.
so u can keep those two qualities in place and then get very, very creative
Barring individual preferences, girls *love* being given a pet duck to take care of and keep alive. It triggers their motherly instincts. No need to wait till the second date, even
Reply to @mealreplacer, 19:25 UTC · 6 likes
(The tweet being replied to by @mealreplacer is no longer available.)
The potential here is endless: tea, nicotine, adderall, desoxyn (aka meth), the mythical *dex*methamphetamine, placebo pills, theobromine, selegiline, l-dopa, tyrosine, Diet Coke, both other kinds of coke, bupropion, parnate, xyrem, maybe even “diet exercise and sleep”
Reply to @ohwizenedtortle, 19:43 UTC · 7 likes
oh dandy lion🔸 (@ohwizenedtortle): Neato!
image attached
(quoting @iamaheron_: well, anon? https://psychology-tools.com/test/autism-spectrum-quotient/score)
🤔
Reply to @ohwizenedtortle, 23:25 UTC · 7 likes
oh dandy lion🔸 (@ohwizenedtortle):
On 6 I was like, “how could you know what you do and don’t usually notice?” You only get to know the numerator!
Reply to @AndyMasley, 23:26 UTC · 1 like
Andy Masley (@AndyMasley): I lean strong yes but qualia kinda mess me up
Yup
Reply to @mealreplacer, 23:31 UTC · 3 likes
(The tweet being replied to by @mealreplacer is no longer available.)
New alignment strategy: replace every adverb with “stochastically” so the AI will know how smart and therefore unbeatable you are
Reply to @AndyMasley, 23:43 UTC · 2 likes
Andy Masley (@AndyMasley): @AaronBergman18 @AaronBergman18 do you wanna set some kind of learning goal with me to figure out qualia once and for all?
Idk how sarcastic ur being but I 100% seriously support a google doc working group
Monday, March 6, 2023
Quote tweet, 01:36 UTC · 5 likes
100%. While I don’t wish suffering on anyone, I encourage everyone to try to consider their takes on ethics, meta ethics, and even consciousness the next time they are suffering
I’m ~certain that ~all papers on these topics were *not* written in such a state, btw. Free alpha!
Quoting George McGowan (@GjMcGowan): Never feel more conscious, alive and in touch with reality than when I'm in pain
Reply, 05:43 UTC · 2 likes
Good, history is overrated
Reply to @ruth_hook_, 05:47 UTC · 4 likes
(The tweet being replied to by @ruth_hook is no longer available.)_
Just one more beta amaloid study I promise it’ll work this time
Reply to @QiaochuYuan, 05:48 UTC · 2 likes
QC (@QiaochuYuan): anyone else here have favorite etymologies? pls share below. i am collecting them
(quoting @dreamyweather: @QiaochuYuan I think "kosmos" has got to be one of my faves. It has a bunch of meanings including order (like, the order inherent in a system), the world (including heaven and the underworld), and decoration. Hence: the cosmos, cosmopolitan, cosmetic)
“Hot potato” (n) the game inspired by hot potatoes
Reply, 05:56 UTC · 2 likes
Chuckled out loud
Reply to @ESYudkowsky, 19:47 UTC
Eliezer Yudkowsky ⏹️ (@ESYudkowsky): I'm not sure that is true, but my God is that profound.
(quoting @amtrpa: @ESYudkowsky Consciousness is when the mask eats the shoggoth)
@AndyMasley what do we think?
Tuesday, March 7, 2023
Reply to @JMannhart, 02:06 UTC · 23 likes
Jonathan Mannhart 🔎 (@JMannhart): @mealreplacer Oh oh! I also have one!!
“If shrimp welfare activists were ACTUALLY serious, they would [totally insane thing that no person believing this would actually ever do].“
*looks over at my giant tank of shrimp in heroin infused water*
Reply to @StefanFSchubert, 02:09 UTC · 7 likes
Stefan Schubert (@StefanFSchubert): Someone is being wrong on the internet
Indeed, but should the reader infer a connection between this proposition and the photo provided?
Reply to @PradyuPrasad, 02:11 UTC · 4 likes
Pradyumna 🇸🇬 (@PradyuPrasad): @ohwizenedtortle @Laura_k_Duffy I can't get my parents to listen to my own podcast 😭😭
Lmaoo
Reply to @ohwizenedtortle, 02:11 UTC · 1 like
oh dandy lion🔸 (@ohwizenedtortle): @Laura_k_Duffy You got your parents to listen to an 80k episode with you? Can your parents adopt me?
Wait but Chalmers isn’t an illusionist?
Reply to @ruth_hook_, 02:12 UTC · 5 likes
(The tweet being replied to by @ruth_hook is no longer available.)_
What positions? These claims aren’t even about biology
Reply to @lfschiavo, 20:01 UTC · 1 like
Larissa Schiavo (@lfschiavo): what if I told you https://www.etsy.com/listing/1371574690/
Who is this lmao
Quote tweet, 20:54 UTC · 7 likes
Endorse. Not all good things are bad for you!
Quoting The Affective Altruist 💛 (@baesrule1): THEY LIED TO YOU! (... about salt)
No but seriously, you can eat as much salt as you fucking want. Here's what happened: research found that high salt intake increases death and disease in those with high blood pressure. Is this true or false: true! (1/n)
Reply to @StefanFSchubert, 21:40 UTC · 1 like
Stefan Schubert (@StefanFSchubert): Anecdotal impressions from the big country in the West (Swedish expression)
https://x.com/jordanwpieters/status/1616969354507161601?s=46&t=zFSMbdq7sUeaJ_kJen_W-g
Thread, continued, 21:40 UTC · 3 likes
Reply to @natalia__coelho, 21:44 UTC · 5 likes
(The tweet being replied to by @natalia__coelho is no longer available.)
Ok I remember forming this opinion a while ago but don’t remember much more than that. So a @natalia__coelho lit review in the opposite direction does radically decrease my certainty, at least
Wednesday, March 8, 2023
Reply to @Richard_Vixen, 02:37 UTC · 3 likes
(The tweet being replied to by @Richard_Vixen is no longer available.)
Pretty sure I’m at high risk of being this type of guy (conditional on gf, so low absolute risk), and from my (imaginary) perspective this just feels like having a bunch of intellectual discussions - I prob won’t know my partner experiences it differently unless they tell me!
Reply to @SpacedOutMatt, 02:43 UTC · 2 likes
Matt 🔸 (@SpacedOutMatt): @AaronBergman18 @Richard_Vixen 10/10 for "conditional on gf, so low absolute risk"
Gonna use this as an example if I ever have to explain conditional probabilities
Quote tweet, 04:09 UTC · 7 likes
Nah sorry, hate to be that guy but optimism already has the memetic advantage
Hope, humor, gratitude, kindness, and generosity are all good and praiseworthy and nearly impossible to overdo…
Quoting Nathan 🔎 (@NathanpmYoung): I think we should post a bit more about what we are optimistic about
Thread, continued, 04:09 UTC · 1 like
But optimism is the knob that adjusts “okness with the current trajectory” and as such should reflect how ok the current trajectory actually is
Thread, continued, 04:19 UTC · 2 likes
And, for better or worse, “say what you’re optimistic about” suffers from the same universal-accessibility as “class, give the person to your left a complement;” the instruction increases everyone’s expectations, and the value of an unelicited positive remark falls
Reply to @CineraVerinia, 09:00 UTC · 2 likes
(The tweet being replied to by @CineraVerinia is no longer available.)
Unfollowing you for a pretty tame critique, not the take per se
Reply to @ESYudkowsky, 14:04 UTC · 14 likes
Eliezer Yudkowsky ⏹️ (@ESYudkowsky): As somebody who's sane but not autistic, it's been very odd to watch the meaning of the formerly technical term "autistic" expand and mutate to mean "sane".
Idk if I’m sane, but gonna start calling myself “autism adjacent” and you’re welcome to join
Reply to @keithfrankish, 14:45 UTC · 5 likes
Keith Frankish (@keithfrankish): What are some good comfort twitter feeds and blogs, for times when you need a break from the stress and worry of modern life?
@ESYudkowsky
Reply, 18:21 UTC · 3 likes
that chopstick was not coming out of my neck on its own (true story btw, 3 year old me was not very good at maximizing his inclusive genetic fitness)
Reply to @QualyThe, 21:09 UTC · 3 likes
Qualy the lightbulb (@QualyThe): yo check out how much of the world's biomass is made up of trees (measured on a log scale)
Is the joke the log scale thing, the fact that trees dont exist, or something else Im too autism-adjacent to get?
Reply to @AndyMasley, 21:44 UTC · 4 likes
Andy Masley (@AndyMasley):
Andy you make IB physics explainer videos
Reply to @AndyMasley, 21:50 UTC · 1 like
Andy Masley (@AndyMasley): I am extremely good at high school physics
Indeed, and this is strong evidence of abnormally high intelligence
Thursday, March 9, 2023
Reply to @daniel_271828, 03:13 UTC · 5 likes
zzDANIEL_ETH (@daniel_271828): I think the most reasonable argument someone could give me for voting Republican is “I’m a single issue voter and my single issue is shutting down gain-of-function research”
(quoting @R_H_Ebright: "Knowing the origin of COVID..is fundamental to..prevent future pandemics." "Mounting evidence continues to show that COVID..may have originated from a lab in Wuhan, China." "The..media downplayed—and..denied—the..theory..COVID..emerged from..WIV."…)
Or it’s 1860
Reply to @daniel_271828, 03:24 UTC · 5 likes
zzDANIEL_ETH (@daniel_271828): @AaronBergman18 Okay, but it’s… not
Good point, updating
Reply to @absurdlymax, 06:18 UTC · 2 likes
Max Alexander (@absurdlymax): @sixers2772 @AFraserUrq I didn't say I wanted to, simply that I was willing
What does this even mean lol
Reply to @pli_cachete, 06:29 UTC · 5 likes
(The tweet being replied to by @pli_cachete is no longer available.)
That better be a natural log or else I’m rejecting mathematics
Quote tweet, 06:41 UTC · 2 likes
🐐
Quoting who cares (@DianaG2772): Legit call from the school:
Principal: I just wanted to touch base with you. Your daughter was baiting seagulls into the playground with gummy worms and actually caught one; Like in her arms. It did bite her—not hard, but I needed to inform you that we filed an incident report
Reply, 10:30 UTC · 2 likes
No no no the Mario arc cannot extend past Waluigi
Tweet, 11:05 UTC · 4 likes
Y’all need seven books? Skill issue
Reply to @ruth_hook_, 20:03 UTC · 2 likes
(The tweet being replied to by @ruth_hook is no longer available.)_
I don’t think it usually literally means that, just like how “first order effects” or “direct effects” aren’t usually appealing to irreducible physics like gravitational attraction
But in general everyone knows what it means in a given case so it’s a useful distinction
Reply to @undo_hubris, 20:15 UTC
Rubi Hudson (@undo_hubris): @storebrandguy If you are a single-issue voter, also don't vote. You're probably a moron too. Oh, you just happened to pick the correct issue to base your vote on, unlike all those other single-issue voters who picked the wrong one?
Yes
Reply, 20:22 UTC · 6 likes
https://x.com/aaronbergman18/status/1624136225106890764?s=46&t=zFSMbdq7sUeaJ_kJen_W-g
Reply to @NathanpmYoung, 20:23 UTC · 2 likes
Nathan 🔎 (@NathanpmYoung): Some things I want:
- Being able to type ideas for web apps and they appear
- To see some live dinosaurs
- For everyone to have their basic needs met
- To live within walking distance of many friends
- To produce a nerdy rap albumWhat do you want?
Nipsey Hussle ft Aaron Swartz 🥲
Friday, March 10, 2023
Reply, 05:32 UTC · 13 likes
I have zero idea what you’re in for, but this basically took over 3 days of my life and for a couple hours I thought I might be like Milo-level cancelled. Muted it, never locked, zero effect on anything whatsoever afaict besides being a funny anecdote https://x.com/aaronbergman18/status/1535968622081867777?s=46&t=zFSMbdq7sUeaJ_kJen_W-g
Thread, continued, 05:33 UTC · 2 likes
Now THAT is a ratio!
Reply to @daniel_271828, 05:42 UTC · 8 likes
zzDANIEL_ETH (@daniel_271828): @AaronBergman18 @ChanaMessinger "thought I might be like Milo-level cancelled"
I just want to say this is an absurd thing to think given what the content of your tweet actually was
Yeah I know, it sounds and is insane lmao. But “infinity people yelling at you on the internet” probably feels pretty similar regardless of whether you did smth wrong and whether infinity means 3,000 or 3M
Reply to @daniel_271828, 05:42 UTC · 2 likes
zzDANIEL_ETH (@daniel_271828): @AaronBergman18 @ChanaMessinger I mean this in a "don't feel bad about yourself – you weren't close to that" kind of way, not a "haha how could you think something so dumb" kind of way
Ya ik haha thanks
Reply to @MagusGenji, 10:59 UTC · 2 likes
magusgenji (@MagusGenji): @cauchyfriend Jokes on you, I created systems for talking to people at parties years ago.
Share👏systems👏
Reply, 14:38 UTC · 2 likes
I wrote a minor thing that seems relevant https://forum.effectivealtruism.org/posts/Xw98osdN4mEGbRDiR/aaron-bergman-s-shortform?commentId=6tTp8KgouKixEzyYv
Reply to @sleight_henry, 14:45 UTC · 2 likes
Henry (@sleight_henry): Rationalists will pick inspiring org names like CFAR or Lightcone and EAs will literally name theirs CGRSP
Lightcone is a pretty sweet name
Reply, 14:47 UTC · 4 likes
But personally I’d prefer Centre on the Global Development for the Future of Post-Human Life Initiative
Reply to @Laura_k_Duffy, 19:14 UTC · 1 like
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): @NathanB60857242 Come join the DC crowd! @SpacedOutMatt @AaronBergman18 @GluskerMartin
My actual best guess (without knowing much more than can be gleaned from ur twitter profile) is that DC is the all things considered best choice. Tbc we’re talking like 35%/45%/20% or maybe even a few points from dc to sf
Thread, continued, 19:19 UTC · 1 like
My sappy poetic altruistic case here is that DC is (substantively, not just de jure) the center of political power for the most powerful institution ever to exist in the universe (the US federal gov)
Reply to @AaronBergman18, 19:34 UTC · 2 likes
(The tweet being replied to by @AaronBergman18 is no longer available.)
- @AndyMasley runs EA DC and so the latter must be good by transitivity of goodness
- ppl disagree with me I think but everything here should be weirder and generally more (on the margin, not all the way!) SF, and probably on avg importing more of EA Twitter would help w this lol
Reply to @SpacedOutMatt, 19:41 UTC · 1 like
Matt 🔸 (@SpacedOutMatt): Representatives from the Ophiuchus Supercluster Galactic Hyperconfederation reading this like 👀
Representatives from the ₽•æœ‰’s Post-Republic of the Hercules–Corona Borealis Great Wall Petastructure reading *this* like 🧿👁️🧿👁️🧿👁️🧿👁️
Thread, continued, 19:43 UTC · 2 likes
STOP doing physics, there should not be 2 kinds of distance
Reply to @NathanB60857242, 23:33 UTC · 2 likes
Nathan Barnard (@NathanB60857242): i'm expecting to be rejected a lot over the next year, this tweet will be the start of a public record of all the things i get rejected from
Saturday, March 11, 2023
Reply to @peterwildeford, 02:51 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18 @frances__lorenz @Mihonarium @RethinkPriors Just wanted to swing back to this and say that @RethinkPriors has definitely been aware of this
(they are pleased)
Reply to @panickssery, 02:58 UTC · 1 like
Arjun Panickssery (@panickssery): An absurd amount of my mana is in this one market on whether a friend will bench two plates by September
I'm not sure if I'm doing this Manifold Markets thing right
What’s his current 1rm?
Reply to @panickssery, 05:39 UTC · 1 like
Arjun Panickssery (@panickssery): 155
If he’s been lifting (really, doing pushups or bench averaging >= 30 reps/week) for more than 6 months I’d bet against, less than a month I’d bet for, probably the 50% line is around 10 weeks but 1-5 mos I’d have to think
Thread, continued, 05:44 UTC · 1 like
Nothing against him! Conditional on caring enough to have a manifold, his 1rm lift gains almost certainly aren’t limited by willpower/willingness to do what’s necessary or knowledge about routines and stuff. Or maybe those mean the difference between like 200lbs and 202.5
Reply to @panickssery, 06:31 UTC · 3 likes
Arjun Panickssery (@panickssery): @AaronBergman18 Disagree: 17 years old, male, linear progress since starting in Sept, if he eats enough protein and sleeps enough and takes the creatine I gifted him and doesn't miss a workout for the next 25 weeks I'd say 90%
So in my mind this is a "will devansh lose the habit" market
$20 to charity of winner’s choice?
Reply to @robinhanson, 07:04 UTC · 3 likes
Robin Hanson (@robinhanson): Who are good people to follow for deep timeless general insights, instead of dunks and quips on the topic fashion of the day?
If you can look past the _obligatory_ conference/event/book/video promotion, @davidchalmers42
Reply to @absurdlymax, 07:06 UTC · 2 likes
Max Alexander (@absurdlymax): @Laura_k_Duffy You seem fairly based to me
Indeed it appears so
Tweet, 13:37 UTC · 10 likes
Ooh fdic discourse. If anybody wants to know how *precisely* the interval between federal on-site safety and soundness inspections is determined for banks with a CAMELS rating of 3 or below, I suspect I was the world’s leading expert on this particular question for a week or two
Thread, continued, 13:37 UTC · 5 likes
at some point in late 2021
…Now you’re gonna have to ask someone else bc they didn’t let me keep any of the things I wrote (security and all ik ik)
Thread, continued, 13:37 UTC · 4 likes
Oh you didn’t want to know that? That has nothing to do with the Current Thing? Ok ☹️
Reply, 21:38 UTC · 2 likes
He just like me fr
Reply to @GhostCoase, 23:44 UTC · 2 likes
Coase’s "Net effect?" Ghost 🏴🕊️ (@GhostCoase): Hmmmm
Obligatory: atheist, hopefully not racist, Jewish enough to have had a Bar Mitzvah in Israel, a couple theology classes in college
Pretty sure “atheism worse” follows in a boringly unambiguous way from the scriptures of at least the Abrahamic religions (if you regard them as
Sunday, March 12, 2023
Reply to @peterwildeford, 06:47 UTC · 7 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): I mean if the only way you know how to find love is to sexually harass people, then I really do think you should not find love - sorry that's the crux
I also don't think banning dating at work is a good idea and I don't see why this is such a binary thing for you.
I think you’re some combo of misinterpreting and being abnormally uncharitable here
Sexual harassment is absolutely a huge problem and not ‘just’ in absolute terms but relative to very many bad things at least as salient in Western politics/media/culture (this is my impression
Thread, continued, 06:52 UTC · 5 likes
rn not based on any recent concrete research), other categories of crime providing some examples esp when weighted by salience (eg gun violence)
But declining quality and number of relationships (sexual/romantic and social, also same epistemic status^) is up there too!
Thread, continued, 06:55 UTC · 2 likes
It’s legitimately unclear to me which is “worse”/more important all things considered (as a straight guy I should say), though I’d probably guess the latter; maybe I’d put 70% on coming to this conclusion after learning a lot more about both
Thread, continued, 06:58 UTC · 1 like
FWIW idk how much of your disagreement leans on the birth rates thing but I haven’t been thinking about that at all and am less sure than most EA adjacent neolib types that at the current margin it’s even a bad thing on net
Reply to @peterwildeford, 08:01 UTC · 1 like
Peter Wildeford🇺🇸🚀 (@peterwildeford): Also stop me again if I'm being uncharitable, but I think I'm bothered by what appears to be "won't women being harassed just think of the declining birth rates!" logic, as if women are just baby-making machines or something?
Mistake theory! (and see below for directionally anti-anti-anti-natalist propaganda)
Thread, continued, 08:03 UTC
I agree with you on the birth thing. This tweet from before doesn’t address the misogyny pronatalism is sometimes a mask for
Thread, continued, 08:09 UTC · 4 likes
New cause area: abolish pronouns (nothing to do with gender/trans issues, they’re just occasionally ambiguous)
time to log off
Reply, 08:15 UTC · 7 likes
(The tweet being replied to is no longer available.)
Guy who does this solely to weed out girls for whom it’s a deal-breaker
Thread, continued, 08:17 UTC
Girl who’s deal gets broken solely to weed out guys who are playing this particular game of 4d signaling-selection chess
Thread, continued, 08:17 UTC · 2 likes
It’s turtles all the way down from here
Reply to @Aella_Girl, 08:18 UTC · 21 likes
Aella (@Aella_Girl): is there any way to dye semen before it comes out?
If this is possible to do in a reasonably safe way, my money is on beta-carotene megadose
Thread, continued, 08:19 UTC · 3 likes
Dude will turn orange everywhere tho, it’s not selective
Reply to @frances__lorenz, 17:54 UTC · 4 likes
(The tweet being replied to by @frances__lorenz is no longer available.)
Agh I strongly disagree that is an implication of or essentially my view! I just need more than 6 minutes for this one which is what I have rn!
Reply to @peterwildeford, 17:56 UTC · 2 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): @AaronBergman18 @frances__lorenz @JMannhart @kevinf567 @emmalrichter @ChanaMessinger @NathanpmYoung I think all three of us disagree with you* btw, I just didn't want to create a pile on.
-
*Or rather - disagree with our interpretation of you, which is different from what you actually meant? Still unclear what you meant but will give benefit of doubt.
Major +1 on the asterisk lol
Monday, March 13, 2023
Quote tweet, 18:14 UTC · 19 likes
50/50 with 700 votes in a de facto referendum on the status of EA among a non-trivial share of the policymaking elite in the most powerful institution ever to exist in the universe?
This is the most important vote you will ever cast
Quoting New Liberals (@CNLiberalism): Poster Play-In (Round 1)
---
Zoning abolitionist @theurbaneist
vs
EA Enthusiast @Laura_k_Duffy
Thread, continued, 18:14 UTC · 6 likes
I typed out “I’m [x]% serious” with x ranging from 35 to 51, so I guess I’m 43% serious
@Laura_k_Duffy, who’s more numerate than me, can check my math
Reply to @BootsPlatypus, 23:18 UTC · 2 likes
Platypussy in Bio (@BootsPlatypus): @AaronBergman18 "policymaking elite in the most powerful institution ever to exist in the universe" we have differing perceptions of ne0liberal i guess
Followed by Ezra Klein, Yglesias, David Shor, Nate Silver, to name a few. A twitter follow only means so much but also the shill thing definitely reaches into wonky policy twitter more broadly
Wednesday, March 15, 2023
Quote tweet, 02:07 UTC · 31 likes
Yeah I feel this super viscerally for the first time literally today. Not "concern," which I've felt before - that emotion is directed at the future, an instinct to fix the broken. To be on the lookout.
This is (or at least feels) about right now
...
Quoting Andy Masley (@AndyMasley): I want to get off Mr. Altman's Wild Ride
Thread, continued, 02:07 UTC · 4 likes
The pic is an ok metaphor for how one sliver of my brain represents part of the current situation, not AI in any other sense!
Tweet, 06:56 UTC · 3 likes
Convince me that brown not being a shade of grey (it’s spelled both ways ok) isn’t *the* glitch in the simulation. Don’t they both reflect all wavelengths??
I can’t tell if I’m mostly tired mostly insane or a secret very weird third thing. Classic fake explanation btw 👇
🦃=🐭
Reply to @mcinthedc, 07:59 UTC · 4 likes
(The tweet being replied to by @mcinthedc is no longer available.)
I will die on the pro-pajama pants hill
Reply to @Laura_k_Duffy, 08:02 UTC · 2 likes
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): @mcinthedc I used to wear pajamas in public all the time, until I met someone who was like “ohhhh you’re the girl who walks around the neighborhood in PJs” 😳
I think we should ban “PJ” - it’s like halfway to “jammie”
Show the garment and it’s neighborhood walking wearers some goddamn respect
Reply to @peterwildeford, 17:47 UTC · 3 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): No need to rush, we will all still be here when you have more time. In the meantime I will give you the benefit of the doubt. I think it would be good to express this well so we understand what you mean.
I might be reneging on this, actually 😔
feels like first smart friend in a math class got x=6 to your x=no solution and inside view, at least, says that they’re wrong
But if you learn your other two smart also got x=6,
(1) all things considered it’s almost certain
Thread, continued, 17:50 UTC · 2 likes
that you’re the one who messed up and (2) it *might* be worth picking through each line to figure out what went wrong, but sometimes it’s not
Here, it doesn’t seem worthwhile now but def could if, say, a bunch of similar issues become action relevant
Tbc I’m saying/now believe
Thread, continued, 17:52 UTC · 3 likes
I screwed up somewhere and at least the point I *expressed* was wrong, regardless of if the thing in my head is true
Reply to @timfduffy, 18:07 UTC · 1 like
Tim Duffy (@timfduffy): @AaronBergman18 I opened up my color research tool (Microsoft Paint) to see what I could find. It seems that brown is just a mix that favors red over blue. For example RGB(75,75,75) is gray, while RGB(125,75,25) is brown.
Ok I’m super confused rn. Maybe slightly validated lmao
Brown: https://en.wikipedia.org/wiki/Brown
Shades of brown: https://en.m.wikipedia.org/wiki/Shades_of_brown
Quote tweet, 18:09 UTC · 7 likes
Guys “the” color brown is very weird
Quoting Aaron Bergman 🔍 (@AaronBergman18): @timfduffy Ok I’m super confused rn. Maybe slightly validated lmao
Brown: https://en.wikipedia.org/wiki/Brown
Shades of brown: https://en.m.wikipedia.org/wiki/Shades_of_brown
Saturday, March 18, 2023
Reply to @Laura_k_Duffy, 09:27 UTC · 2 likes
Laura Muresianu 🌐🔸🇺🇸 (@Laura_k_Duffy): As your Chief Neoliberal Shill, I'll always honor my campaign promises, starting w/my Tipsy Twitter AMA
Post all your qs regarding neoliberalism, MT zoning, chicken welfare, or anything else below! I'll have a few drinks and start answering them at 7
And continue voting for me!
image attached
(quoting @CNLiberalism: The Maia Mindel Region (Round of 32) --- (1) 2022 Shill Bracket Runner-up and tax policy economist @wwwojtekk vs (9) Peter Singer Stan Account @Laura_k_Duffy)
P(run for office)
P(run for MT senate)
What question were you hoping to get asked but didn’t yet?
Ok fine that one^
The cursed version of that^^ if and only if you want to answer
Tweet, 09:53 UTC · 94 likes
Pretty sure but lmk if it seems wrong:
no LLM has generated a correct but otherwise unknowable statement
Stronger version: Arbitrarily large LLMs trained and run in qualitatively similar way to those live rn will be likewise incapable of doing this
Reply to @moreisdifferent, 13:04 UTC · 2 likes
Dan Elton 🔍 (@moreisdifferent): @AaronBergman18 You mean previously unknown?
No, though intuitively that’s the same idea. Eg answer to a lengthy but simple arithmetic problem or “how many instances of the letter u are there in all my tweets?”
Reply to @benrayfield, 19:46 UTC · 5 likes
Lambda Rick 🏴☠️/acc (@benrayfield): If its "unknowable", how would u know its "correct"?
I’m not actually certain this is always formally right but intuitively, by difficulty :
(True) hypothesis generation> proof > verification of proof
But my real answer is “things that look doable only with new knowledge get done” - we may not know what that knowledge is!
Reply to @goth600, 19:56 UTC · 2 likes
@goth (@goth600): Beginning to think even ASI won’t be able to help Sam Altman find cool shoes to wear
If I were him I’d go harder on the counter signaling. Pajama pants. Two different flip flops. A big ziplock bag
Monday, March 20, 2023
Reply, 08:13 UTC · 16 likes
Solve the alignment problem
Thread, continued, 08:13 UTC · 4 likes
Ha u guys that was easy
Reply to @peterwildeford, 22:27 UTC · 3 likes
Peter Wildeford🇺🇸🚀 (@peterwildeford): File this one under obvious. I wonder if we could've made money on this in retrospect? I've never bought or sold an individual stock but some things seem pretty clear?
I don't understand markets. (That's why I don't buy sell individual stocks.)
(quoting @AlecStapp: In January, Buzzfeed stock spiked nearly 300% when the company announced plans to use ChatGPT for content creation. It, uh, didn’t last long…)
Short selling means you’d be on the hook for an unbounded amount of money if wrong
Idk what the options prices were but probably very unfavorable to those trying to do a lower bounded-risk version of this
So even in hindsight it feels very sketchy
Tuesday, March 21, 2023
Quote tweet, 06:25 UTC · 3 likes
Retvrn to 2 parameters
Quoting @goth (@goth600): gm
Tweet, 06:50 UTC · 4 likes
This would be one of the funnier ways to become a homeowner in a $1M+ house neighborhood, and I’m not even sure it’s a bad idea
Tweet, 11:48 UTC · 20 likes
I have no idea what a GPU is at at this point I’m too afraid to ask
Tweet, 20:58 UTC · 7 likes
New LLM just dropped
Wednesday, March 22, 2023
Tweet, 04:19 UTC · 5 likes
Every time I check on my $18 UBOpenAI thinking it should be used up, the little bar thing has moved like two cents
I have $17.48 remaining
Reply to @absurdlymax, 13:02 UTC · 2 likes
Max Alexander (@absurdlymax): What's the cheapest place to buy digital storage?
(Maybe also most secure places as well?)
No superlative claims but I have backblaze
Reply to @moskov, 17:34 UTC · 27 likes
(The tweet being replied to by @moskov is no longer available.)
Domain land value tax when
Thursday, March 23, 2023
Reply to @baesrule1, 03:56 UTC · 2 likes
The Affective Altruist 💛 (@baesrule1): How do you make brown rice taste better?
@Laura_k_Duffy @ryancbriggs
Remove the germ and fiber
Quote tweet, 05:22 UTC · 2 likes
Many are saying
Quoting Malcolm Ocean 🏴☠️ (@Malcolm_Ocean): psych has a context crisis, deeper than a replicability crisis
you can measure people via surveys but what you measure is only how they respond to surveys
very trippy demo (correlations in blank likert scales)
https://x.com/literalbanana/status/1638387396009140224
Tweet, 21:05 UTC · 83 likes
“AI will never-”
“6 months”
Ok wise scale maximalist, what do *you* think AI can’t do in, say, 2 years? Has to be possible in principle tho, ideal probability range like 2%-25%
Thread, continued, 21:05 UTC · 25 likes
I’ll go with “predict large earthquakes early enough in advance for people to leave” and “state what color screen a human is looking at using only neural data”
Tweet, 21:15 UTC · 23 likes
If it turns out these have already been done this was a clever ploy to illustrate the pace of progress
Reply to @QualyThe, 21:25 UTC · 2 likes
Qualy the lightbulb (@QualyThe): you can just send normal AI risk explainer articles now, you don't need to send people HPMOR or the sequences 🥰
What normal articles are ppl referring to? I’ve lost track of what counts as normal
Friday, March 24, 2023
Reply to @undo_hubris, 03:55 UTC · 15 likes
Rubi Hudson (@undo_hubris): Create videos at least a minute in length based on prompts, which are consistently difficult to distinguish from human-generated videos.
Intuitively this seems like 70% likely for casual layperson viewing but idk what im talking about
Reply to his own earlier tweet, 19:16 UTC · 5 likes
Aaron Bergman, March 23, 2023: If it turns out these have already been done this was a clever ploy to illustrate the pace of progress
Update: failed to follow own instructions
Reply to @BenHayum, 21:27 UTC · 3 likes
(The tweet being replied to by @BenHayum is no longer available.)
Do this and I'll stop stanning MS Word. Don't say you weren't warned @MSFTResearch @Office @Microsoft
Reply to @derekofmilford, 21:28 UTC · 2 likes
derek (@derekofmilford): There is no world where zero people pursue this
But there is a world in which a handful of top companies with the most talent and compute don't
Saturday, March 25, 2023
Tweet, 04:56 UTC · 6 likes
I'm testing chatGPT's ability to predict the number of likes, retweets, bookmarks, and comments on this very tweet
Tweet, 08:01 UTC · 1 like
(gif, plays on X)
Quote tweet, 08:12 UTC · 9 likes
Thx
Quoting Aaron Bergman 🔍 (@AaronBergman18):
Sunday, March 26, 2023
Reply to @xriskology, 02:24 UTC · 6 likes
Dr. Émile P. Torres (they/them) (@xriskology): Content warning!
I feel physically ill watching this. "Homo sapiens," the self-described "wise human." Can't believe we treat other creatures like this.
(quoting @goth600: Man made horrors)
We don’t agree on much but we definitely agree on this
Tweet, 22:07 UTC · 3 likes
@tobyordoxford when @lexfridman drops “We stand on the precipice…” 30 seconds into his interview with @sama
(gif, plays on X)
Reply, 22:53 UTC · 12 likes
“Fruit bad actually” is an attractively contrarianish take that makes a good bit of theoretical sense but afaict basically all empirical research contradicts
Monday, March 27, 2023
Reply to @PradyuPrasad, 01:24 UTC · 5 likes
Pradyumna 🇸🇬 (@PradyuPrasad): @peterwildeford > smart country
And what may have given you that impression
I suspect a disproportionate amount of time spent in Oxford
Reply, 14:49 UTC · 2 likes
My sister and I have our own semi-complex language! Ok maybe dialect is the better word, but it’s substantially its own thing lol
Reply to @absurdlymax, 16:02 UTC · 2 likes
Max Alexander (@absurdlymax): Should I thank Qualy in my acknowledgments section?
I managed to miss the voting window but yes this is morally obligatory
Reply to @PradyuPrasad, 16:14 UTC · 2 likes
Pradyumna 🇸🇬 (@PradyuPrasad): I can of course understand the reason for not having it to walk around. But voting?? Ideally you'd want only eligible voters to vote at only one booth!
This makes total sense in theory but zero in the IRL United States where the type of fraud this would prevent seems extraordinarily rare and there are a gazillion political science papers showing it disproportionately inhibits turnout of the least privileged along every dimension
Reply to @absurdlymax, 16:15 UTC · 2 likes
Max Alexander (@absurdlymax):
There are so many appropriate gifs this is but one
(gif, plays on X)
Tweet, 19:36 UTC · 36 likes
Wanna be clear before I drop the $20
If I get chatGPT Plus, do I automatically get access to gpt4?
Tuesday, March 28, 2023
Reply to @austinc3301, 06:02 UTC · 1 like
Agus 🔸 (@austinc3301): I kinda wish that scope sensitivity had a more poetic name to fit with the other two
If only we had some very creative sand…
1. Limitless vision
2. Boundless benevolence
3. Infinite empathy
4. Endless perspective
5. Expansive excellence
6. Unbounded understanding
7. Infinite insight
8. Limitless compassion
…
Thread, continued, 06:02 UTC · 1 like
9. Cosmic consciousness
10. Boundless perception
11. Limitless heart
12. Universal empathy
13. Infinite care
14. Unfathomable compassion
15. Endless benevolence.
Thread, continued, 06:04 UTC · 2 likes
Actually none of these are amazing, it’s having trouble combining the concepts of proportionality and unboundedness
Quote tweet, 14:00 UTC · 3 likes
By induction, the difference between a significant difference between two degrees of significance and an insignificant difference between two degrees of significance is not itself statistically significant
Quoting Jake M. Grumbach (@JakeMGrumbach): some days don't you just feel like shouting this from a rooftop
(ungated: http://www.stat.columbia.edu/~gelman/research/published/signif4.pdf)
Thursday, March 30, 2023
Quote tweet, 04:51 UTC · 3 likes
If a bunch steps in my hyperspeculative amateur reasoning chain check out, you’ll now be able to sensitize your mu opioid receptors and increase your hedonic baseline, if ever so slightly, with stuff you can buy on Amazon👍
(More below)
Quoting Ivan 🔸 (@FrostedSapling): https://www.fda.gov/news-events/press-announcements/fda-approves-first-over-counter-naloxone-nasal-spray?mc_cid=7b1c907bb9&mc_eid=6421b4abc9 OTC naloxone has been approved by the FDA. This is very cool
Reply to @ruth_hook_, 07:42 UTC · 13 likes
(The tweet being replied to by @ruth_hook is no longer available.)_
Wait this is totally reasonably and probably correct. The particulars of AI risk mean that “world end” isn’t (necessarily or generally) a flavor of “hellscape” but more just everyone’s dead - it’s morbid but this probably isn’t a good reason not to have kids
Reply to @JMannhart, 07:53 UTC · 2 likes
Jonathan Mannhart 🔎 (@JMannhart): @Noahpinion There is *such* a long list of very smart people who are incredibly concerned. Neither the “smart“ nor the “very concerned“ is debatable (in good faith).
You can disagree of course, but do it in good faith. With arguments! Good criticism is great!
But this? It's embarrassing.
Yea - although IMO this meme is shorthand for a bundle of arguments or at least implicit propositional assertions (I apologize for writing this phrase) along the lines of “simple heuristics usually work, this is one such case, disagreement is failed intellect signaling”
…
Reply to @chronononono, 07:55 UTC · 3 likes
chrono (@chronononono): @AaronBergman18 @ruth_hook_ There's nothing reasonable about thinking the world will end in 5 years
Hope ur right
Reply to @QualyThe, 08:28 UTC · 4 likes
Qualy the lightbulb (@QualyThe): Shutting down nuclear weapons production for 6 months is dangerous and irresponsible. The excess fissile material produced by reactors in this time will create a "plutonium overhang" making it easier for rogue states to catch up
If I’m reading this right I don’t think it’s a good analogy: Sentence 2 is neither here nor there but if nukes were just invented and there was a multipolar arms race starting up I do think it would be bad for the US to unilaterally stop after making 2
Reply to @AISafetyMemes, 22:44 UTC · 4 likes
AI Notkilleveryoneism Memes ⏸️ (@AISafetyMemes): @AaronBergman18 @QualyThe This is the most EA response to a joke
🏆
(gif, plays on X)
Reply, 22:53 UTC · 1 like
I guess depends on if this means “try to hit location 10k miles away and the st dev of its landing is more like 100 miles, more like 1000, or “sometimes detonates randomly even in storage”
Thread, continued, 23:04 UTC · 1 like
Anyway maybe a better analogy is “fossil fuel reserves are 1/10 their actual size, we just discovered fission but nobody knows how hard reactors are to build safely and if one ever melts down the atmosphere ignites. What should the govt do both legislatively and internationally?”
Friday, March 31, 2023
Reply to @NathanB60857242, 04:47 UTC · 2 likes
Nathan Barnard (@NathanB60857242): feeling 22
Same (neither 22 nor my bday tho)
And happy birthday!
Reply to @mealreplacer, 12:36 UTC · 8 likes
(The tweet being replied to by @mealreplacer is no longer available.)
Quote tweet, 18:31 UTC · 18 likes
Idk who needs to hear this but *there is no legal obligation for corporations to maximize shareholder value* - at least in the US and UK
Quoting zzDANIEL_ETH (@daniel_271828): One little-known fact about the US legal system is that no one actually has a fiduciary duty to kill all of their shareholders. In fact, you could even make an argument that there is generally a fiduciary duty to *not* kill all of your shareholders.
Reply to @cauchyfriend, 20:17 UTC · 2 likes
shako (@cauchyfriend): a few times i've made a reference to twitter IRL and people asked me for my handle, and I had to go "haha... n..no" "wait seriously" "y..yeah, n..no. sorry"
Coworker at a previous job let on that he was in the postrat sphere somewhere but on “deep background” or smth so refused to say more. If that’s you, hope all’s well!
End of March 2023. Previous month: February 2023. Next month: April 2023. Back to the archive index.