Sichu Lu([email protected])
@lu_sichu
Followers
2K
Following
79K
Media
5K
Statuses
51K
nlab fan account, arxiv surveyor, pubmed enjoyer, two culture bridger, vacuous high gossiper, dearth of any domain expertise, reluctant g theorist, gpu poor,
dumb random walk in L2 norm:
Joined May 2021
"Sichu, it's not my fault you haven't read all of the nlab" .Challenge accepted. Belated New Years Resolution.
8
5
47
THIS has to be the greatest washington post headline ever
93
1K
18K
I'm begging mathematicians to think about something other than math
43
130
2K
since we are all swifties today my favorite taylor meme
19
91
914
@kitten_beloved Since I disagree, where is the inequivalence in this? why are male teenagers more capable of giving consent than females? Do we have data showing that consensual sex between older females and teenage boys is "better" than the reverse situation in terms of outcome?.
26
3
468
I was today old when I learned the Iliad is actually suppose to be a eight part series and we lost the rest besides the Odyssey. How have I never encountered this factoid before.
8
50
367
babe wake up new type of manifold has dropped, this is the most horrifying thing I've ever seen who chose this image
12
45
356
@cremieuxrecueil would you agree that catching social contagion easily is a form of mental illness?.
6
3
315
@primalpoly I feel like this self selection lol. The people with productive labs/research probably don't want admin responsibility and just want to get on with their work.
11
7
299
@tracewoodgrains @BasedBeffJezos kinda of weird that I respect his real life ID much more than his alt ego on here.
5
2
211
@alz_zyd_ i remember taking improv speech events in hs and I was always like wtf am I going to talk about for seven minutes when presented with a random topic and amazed that other people just yap about nothing.
4
2
209
is this real??????.
A method to unilaterally disable all nuclear bombs on Earth, remotely and without countermeasure:.It uses a 1000 TeV muon->neutrino beam to penetrate right through the Earth and decay near fissile material, forcing it to 'fizzle' and become useless.
20
10
212
@kylebrussell The real humor of this is that this is probably a superior model to the system we have.
4
0
183
sleep deprivation helps me find the funniest thing on scholar
5
19
188
why don't people teach topology in like high school/middle school, incredibly useful and has applications everywhere in the sciences.
24
12
175
if I have to read another dunk on Murray from mathematicians that are at least >3sd in math ability saying wait" everyone around me can learn math, this one time I taught this kid struggling with analysis". the complete lack of self awareness here.
12
8
163
someone buy this and opensource it for the love of genomics research, you can easily strip out identifiable data(well to the extent you can seperate anyone from their dna anyway) nah fuck privacy let's mad science this one.
15 million people trusted 23&me with their DNA. Now the company is almost bankrupt, the stock about to be delisted, and almost the entire board has resigned. Who will they decide to sell your info to?. 1/
7
9
164
this is a general pattern if you read older literature, there are indicators that a lot of people knew a lot of things that they mention in footnotes/passing/comments to other people that end up as 100 page papers/phd thesis/award winning results.
5
5
153
3.2k bookmarks on a screenshot is hilarious on a meta level!
1
2
145
but how is it possible that people at a conference for education are not aware of this? given this gap is reproduced in ALL THE EDUCATIONAL METRICS WE MEASURE.
holy shit. is it possible they just didn't know?.
8
1
141
@tao_lin "At this rate, the majority of American boys will be autistic by 2036, and by around 2045 most children here will be nonverbal." the power of linear extrapolation.
5
2
126
philosophers where are your qualia now
this reads like an amazon wishlist of papers from me two years ago
15
18
128
I was looking for this meme and it was on my phone the whole time. no i don't :(
3
5
126
coward. when you are an assistant professor of statistical genetics and you block some rando on the internet for checking your claims about papers that you keep getting wrong.
9
3
109
good thread, this dude is a state representative and he reads papers about chemical pesticides? based.
1/. I did a deep dive on glyphosate a few years back. Had a friend who pointed me to a site listing ~30 papers. I downloaded and read every single one. One or two made actual claims. but the n and p values were absolute garbage.
6
5
104
this is the like the perfect google search response, first stack exchange, then lesswrong, then IEEE paper, and then wiki
5
2
97
all this talk about AGIs going foom and no one talking about this classic text.
7
5
89
@lymanstoneky I feel like this is sort of predictable from what we know about welfare programs already, there are good ways of giving people grants and also this probably increases net happiness anyway.
3
0
91
@RexDouglass but what was your prior on him being able to understand the bayes calculator???.
0
0
89
Absolute insanity, apparently Russia was shelling civilian target because they tore down old Soviet era military installations and converted to apts and the Russians were using old maps.
7
2
88
@Bent0916 @eleanordotcomm it broke mine first when it kept refusing to answer open research questions.
1
0
81
still one of the most insane things i've ever read.
9
8
86
@1owroller it's a shit test, don't trust it. just take gre/sat/act/mcat/lsat scores and check the correlation.
8
1
85
fomenko does math, crazy historical theories and art? truly a polymath of the highest caliber.
5
3
81
@alz_zyd_ yeah urban texas public hs. I remember one guy from my year that got like awards at like the state level and he was incredibly smooth at saying nothing for seven minutes lol.
1
2
85
should I feel bad for Chomsky. that first sentence.
3
12
87
AGI discourse is making me want to quit twitter for a bit. The general competence of everyone is awful and low and this includes a lot of senior people that should know better. On every possible side of this issue. Schizo anons seem to be the ones properly calibrated ironically.
7
5
82
okay guys this is literally my feed live rn and it's too much pls stop posting about this dude
7
3
82
Probably the most interesting part where they explain why mcts is hard in token space
4
6
81
Hang on I'm on board 99 percent with this but why do they think the other frontier labs are super behind on this? If anything the margins are narrowing for small/open/other companies they are catching up faster/already sota.
Gwern on the recent change in tone, and on Altman's shrinking timelines.
7
2
76
@SkotiVi @kitten_beloved If you are asserting something that requires empirical data then I expect to see data, it's simple. This is just cope. I'm not asking for evidence that 1+1 =2.
2
0
68
As stupid and nitpicky as this particularly is, if we are going to scrutinize academic works for plagiarism at this level that's a net good because god knows academia is terrible and full of frauds.
I was forwarded an email this morning from a reporter at Business Insider who noted that there are four paragraphs in my 330-page PhD dissertation: “Material-based Design Computation,” which I completed at @MIT in 2010,. where I omitted quotation marks.
5
6
72
ladies, and gentleman, this is how you win the nobel prize in medicine. bravo Werner Forssmann
4
4
73
@johnloeber saving those precious pdfs for a rainy day huh. i have a couple of fiction books i haven't read for this reason.
1
0
71
the countersignaling move is to ignore all of that and tweet as your heart desire, bicker and dunk as often as you want, make friends with small accounts cuz they are cool, replyguy to big accounts that don't see you and ignore you.
Twitter dynamics are so fascinating. Smaller accounts tag big accounts, often with ingratiating comments meant to curry favor, and get bits of attention in response. People “invest” in small accounts that will make good allies in the future. etc. First time I’m understanding what.
6
2
72
okay so this monetization thing how do I do this. okay i know I will post cute pictures.
10
4
68
@eigenrobot the phenomenology of spirit is an attempt to talk about dynamical systems by someone with no math training and lack the proper formalism to do so. no further questions.
4
2
61
@captgouda24 On one hand it sucks being the smart kid who don't have same aged peers interested in the kind of stuff they are into, but on the other hand as an adult like hell yeah you want to avoid any improper entanglements with minors.
3
0
61
@eyeslasho Is there evidence that being more selective is necessarily leading to smarter kids.
3
0
57
@pli_cachete strange meta phenomenon where some of the best informed people on abstract theology/apologetics are not believers.
3
0
59
@tracewoodgrains see and also scott's comment/post on this.
Does fluoride in the water supply make you dumb?. Water fluoridation is intended to improve dental outcomes, but there are those who believe that it's part of a conspiracy to make people weak, to make people dumb, or even to make people obedient, among other things. Advocates of
4
1
59
can't believe I have never heard of this before.
1
12
57
@goth600 nah nah goth mine is sexier by far
okay so this monetization thing how do I do this. okay i know I will post cute pictures.
3
1
50
@HyperboIeva I can see this future where smart kids learn about machine learning in a general way before they touch multivariable calculus.
2
0
54
speed/accuracy tradeoff in genes??? some cs algo shit going on here.
3
4
54
@pastaraspberry @arithmoquine that's fair! but in principle you can get the person to say *a sentence has never been uttered before* and etch it with a scanning tunnel microscope that's pretty legit!.
3
0
55
are you guys tired of winning yet jesus what is this the third/fourth model from a chinese team? either anthropic/meta/deepmind/openai has been holding out on us significantly or they are getting their ass kicked.
ByteDance announces Doubao-1.5-pro. - Includes a "Deep Thinking" mode, surpassing O1-preview and O1 models on the AIME benchmark. - Outperforms deepseek-v3, gpt4o, and llama3.1-405B on popular benchmarks. - Built on a MoE architecture, with activated parameters far fewer than
7
4
56
von neumann is way too well known to be the hipster's choice.
My hot take is that Grey Tribe weirdos ultimately like Von Neumann because Einstein is famous, and so by saying "I know the secret, smarter guy", you get to perform some kind of esoteric Smart Guy knowledge. It's intellectual hipsterism for compulsive contrarians.
11
2
56
@coldhealing This is just because the verbal scores for STEM people are generally just as high if not higher than humanities people on average. there is a nice chart of this I have to find.
2
1
49
@joftius maybe you should be more charitable xd
A typical effect size for a successful intervention in education, run and evaluated very well, is about 0.06 d. Low-quality papers suggest huge effects, but they don't hold up to scrutiny.
1
0
55
@knrd_z Maybe we should redefine expertise, so that you are really are expert when no anon with a cartoon pfp can own you.
5
2
50
people do not realize that their opinion of the world is also a confession of character.
Aella and Eliezer see relationships as transactional with partners always looking to trade up to better models. Which is ok, for them. What's strange is Aella's insistence that everyone else must be the same but in denial!. Not so! We humans see romance in many different ways!
4
5
50
dying. imagine going to a prestige business school to network(otherwise why are you there) and missing the future PM of britain.
All the Big Tech PMs that ignored Rishi during his Stanford MBA are gonna have hell to pay at the 20th anniversary class reunion in 2026
0
4
52
@Aella_Girl even massive population sized databanks(such as the uk biobank) suffer from bias and ascertainment issues and people point this out all the time i don't think this issue is unique to the your surveys.
3
1
52
@katiedimartin Sigh, okay what level of evidence would you need to prove 1. g is real and heavily influences life outcomes. 2. there are racial disparities. I'm more than happy to engage/provide providing it is in good faith and don't turn into a screaming match about racism or something.
1
0
46
i like the math as occult knowledge aesthetics but almost nothing else about this analogy really works. math is not hidden to anyone, you don't technically need people to teach you anything if you are smart enough, it is open to you with pen and paper.
cheesy as fuck to say, but that occult knowledge you're looking for IS real and it's pure math. like, the esoteric teachings that grant you sight beyond the veil to witness the underlying machinery of reality EXIST and ARE HIDDEN, just not by conspiracy but by their own nature.
8
0
49
@katiedimartin this is funny but I think most of this material is aimed at people with out linguistics backgrounds and find different things to be "basic".
2
0
49
I'm pro AI slow down because I ain't got time to read arxiv papers.
3
3
45
@littmath sorry i am more confused now, it seems like the pisa scores and world bank learning rate data being cited to compare against lynn's imputation are completely independent. where are you getting this from?
2
0
41
another variant"but i struggled with calculus in hs and now I have a PHD in algebraic number theory".
1
0
42