@niplav | show @replies
[759 updates]
# run this to follow @niplav in the app
[6 days ago]
EXPLAIN GAUSSIAN PROCESSES TO ME OR I'LL FUCKING KILL YOU I'LL WHY THE FUCK HOW THE FUCK CAN IT BE A DISTRIBUTION OVER FUNCTIONS SO THAT THE JOINT DISTRIBUTION FOR OUTPUTS IS GAUSSIAN
[6 days ago]
The Computing Amish vs. Computing Scientology: The Final Showdown
[6 days ago]
snicean relevance gnorms
[7 days ago]
petasus, petasī: not a type of horse
[7 days ago]
disappointed beyond measure that payer doesn't mean “to country” in french
[7 days ago]
the hindley-milner algorithm
[7 days ago]
the cocke-younger-kasami algorithm
[7 days ago]
the bron-kerbosch algorithm
[7 days ago]
schein mir in die augen, kleines
[7 days ago]
degenewussy
[7 days ago]
if something is strictly more intelligent than you, you can't predict how much more intelligent it could be
[7 days ago]
A: “i should be stealing twin babies and giving them up for adoption to advance social science.” B: “fraternal or identical twins?”
[8 days ago]
heterotic string theory pun, knock yourself out
[8 days ago]
(after receiving lessons for ~14 years, but that's not important right now)
[8 days ago]
i am awesome because i can play the Piano
[8 days ago]
cities lack places where you can go and scream very loudly
[8 days ago]
for the love of god i really hope that insects can't suffer
[8 days ago]
[9 days ago]
mARLEBCAKEALSOTHEGAME
[9 days ago]
The stories and information posted here are artistic works of fiction and falsehood. Only a fool would take anything posted here as fact.
[9 days ago]
my kink is people calling me “anon” during sex
[9 days ago]
You have no chance to update make your time.
[9 days ago]
all your bayes are belong to us
[9 days ago]
the awareness will continue to expand until the morale improves
[9 days ago]
contemplating what i must have done right to end up in this moment
[9 days ago]
my problem with animism is that it's too anthropocentric
[10 days ago]
i wonder whether this is to teach us about the r-wordedness of data in the real world.
[10 days ago]
professor gives us data in a dxn ndarray, where d is the number of features and n is the num of datapoints. r-worded
[10 days ago]
[coherent & sensible screaming]
[11 days ago]
non-linked & linked back footnotes in pdfs are annoying
[11 days ago]
gah, can't snarf evince
[11 days ago]
this holds for mathematics as well
[11 days ago]
i think that every language should have a unary - and + operator, and that those should be idempotent. sign flipping should be done by multiplication with -1.
[12 days ago]
kuracisto estas racisto kiu kuras
[12 days ago]
my nervous system is as smooth as a baby
[12 days ago]
the repugnant conclusion turns around: “It's too full here, let's go home.”
[12 days ago]
a hedonium shockwave, a utility monster and the repugnant conclusion go into a bar.
[12 days ago]
synthesis: the reason why african americans commit more crimes _is_ genetic, but responsible are the genes from the white slavers in their ancestry
[2 weeks ago]
reminder to self: clanishness vs. tribalism
[2 weeks ago]
also no simple latex macro for the symbol for conditional independence ⫫
[2 weeks ago]
doesn't evoke the image of being wrong to me, much rather some equanimous exploration
[2 weeks ago]
beautiful though is that errare meant “to wander, to stray” originally
[2 weeks ago]
“ut: like, as” fsdjkljiasdasdf
[2 weeks ago]
the IPA for tous bas (“very small/silent”) is /tu ba/ :-D
[2 weeks ago]
“ainsi: thus, so as.” yeah. hm. i will definitely remember _that_
[2 weeks ago]
every romance language has like three different words for “as”, “like” and “how”, and i can't tell them apart FOR THE LOVE OF GOD
[2 weeks ago]
ĉiu volas vojaĝi al la aja, neniu ŝatas la mala versiono :-/
[3 weeks ago]
i also want to remember 𖤶
[3 weeks ago]
from now on, i will use  whenever abbreviating the Fast Fourier Transform
[3 weeks ago]
help i'm under distributional shift
[3 weeks ago]
i'm a self-professed sexist, but i am not a sexist
[3 weeks ago]
multimodal languages (subjects are spoken, verbs are gesticulated, adjectives & adverbs are grimaced)
[3 weeks ago]
enjoy the once-in-a-while outages of schelling.pt, gives my brain room to think :-D
[3 weeks ago]
Being being been.
[3 weeks ago]
antinatalists: another victim of the replication crisis
[4 weeks ago]
shedding, shedding!
[4 weeks ago]
well, what if my values include a constraint that my values shouldn't be optimized by an outside optimizer?
[4 weeks ago]
gnonomics
[1.2 months ago]
valley of bad rationality girl
[1.2 months ago]
your incentives are bad, and so are you
[1.2 months ago]
“spiritual but not religious” <=> “consequentialist but not utilitarian”
[1.2 months ago]
cats are better than dogs because, well, have you ever stepped in catshit?
[3.1 months ago]
walking through carefully calculated rivers of blood
[3.1 months ago]
fucken double-binds
[3.1 months ago]
otoh i don't trust gov to do pigovian taxes well either
[3.1 months ago]
everybody is an egg for buddha nature
[3.1 months ago]
at least waitt
[3.1 months ago]
wagmi (or, unfortunately, nigmi)
[3.3 months ago]
we have the fermat polygonal theorem and pollock's polyhedral numbers conjecture. can we generalise this into the n-polytopal conjecture?
[3.3 months ago]
how _do_ you prove that n=Δ+Δ'+Δ''?
[3.5 months ago]
“you guys are getting added to groupchats?”
[3.5 months ago]
i innoculated myself against emacs by first & only playing tetris in it
[3.5 months ago]
The Racist in the Cave
[3.5 months ago]
Niplav Yushtun™: A Large Variety of Extremely Good Posts
[3.5 months ago]
I need to find evidence for/against the claim that there was a training run of GPT-2 that maximized negative log-loss – I've heard it a couple of times on the internet and already spread the meme myself, but I haven't seen it in a paper or blogpost
[4.2 months ago]
@movq "How does one interpret those numbers? Does it mean that people usually died at about 30 years or is that really an average, meaning lots of children died but those who survived still reached something like...
[4.3 months ago]
mathematics is the science of the infinite, computer science is the science of the finite
[4.3 months ago]
if you tickle a sub, is the result a bottom contraction?
[4.5 months ago]
what kind of job does one do if one is exceptionally motivated to (and reasonably good at) organising physical objects?
[4.5 months ago]
alternative formulation of orthogonality hypothesis: the size of the space of possible goals for an agent correlates positively, rather than negatively, with an agent's optimization power
[4.5 months ago]
note: the previous twt does _not_ assert the equivalence of these ideas.
[4.5 months ago]
The Orthogonality Hypothesis is obviously true, the Factored Cognition hypothesis is obviously false, and who can tell about the Fundamental Theorem of Finite Factored Sets?
[4.5 months ago]
learning pickup is rewriting your life-history strategy
[4.7 months ago]
schelling.pt down :-(
[4.7 months ago]
a pride flag for heterosexual men who don't like women
[4.7 months ago]
what is the value of knowing the action of a more powerful optimizer in a single universe-state?
[4.7 months ago]
kind of insane that the average life expectancy in 1800 was nearly half as much as the one in the country with the lowest life expectancy today (28.5 years then, 50.7 years today in Lesotho)
[4.8 months ago]
limited wireheaders seem safe but useless (modulo experimentability)
[4.8 months ago]
one might argue that we have huge amounts of uranium and other radioactive elements in the earth's crust already, and since we only extract energy from nuclear energy, putting the waste back should make the earth less dangerous, not more. but the isoto...
[4.9 months ago]
@adi "@niplav I got the joke!" -> Congrats! i didn't
[4.9 months ago]
every wisepost can (but doesn't have to) be credited to carl shulman
[4.9 months ago]
yes, I'm in favour of a combination of fusion power and solar (dyson sphere)
[4.9 months ago]
father (pointing at night sky): one day this all will be yours, son.
[4.9 months ago]
really mad about the death of the FAQ
[5.0 months ago]
gigapearce
[5.0 months ago]
we probably have a kilofoq of value in this universe, probably more (https://arbital.com/p/cosmic_endowment/)
[5.0 months ago]
foq: (ten to the) fifty-one QALYs
[5.0 months ago]
bigword bigwords bigword2
[5.0 months ago]
joule supremacy
[5.0 months ago]
guy who can't use a twitter account responsibly
[5.0 months ago]
age of em is transpositive send twt
[5.0 months ago]
just to make clear: people who actually identify as attack helicopters are unironically valid
[5.0 months ago]
the hardest sci-fi is just 22 pages of differential equations describing the dynamics of Omohundro sludge
[5.2 months ago]
meme innoculation
[5.2 months ago]
Die Wimpern der Cherubim
[5.4 months ago]
I want you to spell out the context. like in porn
[5.4 months ago]
[5.4 months ago]
maybe elites appear incompetent because they're mostly trying to prevent AI disaster behind the curtains? perhaps that's why GPT-4 hasn't appeared yet – they coordinated not to do it.
[5.4 months ago]
on the other hand, induction has also never worked before for anti-inductionists, so they might try it.
[5.5 months ago]
lol, just realised that standard AI risk is just environmental disaster risk amped up to 10^(10^10), only that the direct perpetrator isn't humanity (we're only the proximate perpetrator). huh
[5.6 months ago]
at some point you realise that david pearce is right about everything, nick cammarata is right about everything, andrés gomez emilsson is right about everything.
[5.6 months ago]
[5.6 months ago]
what,,, you are trying to do good? how about I complain about you wanting to do good, and accuse you of being evil, even though on further questioning I would agree your actions are at worst net-negative?
[5.6 months ago]
children are also just self-replicating probes
[5.7 months ago]
not a big fan of retributive justice, but considering the fact that mosquitoe-borne diseases have killed **half of all humans that have ever lived**, we might consider exterminating the fuckers just because they deserve it.
[5.7 months ago]
people who use kilowatt-hours of something should go f**k themselves
[5.7 months ago]
collecting large STEM datasets for training huge models a good step for avoiding human models?
[5.7 months ago]
beating the ender dragon in minecraft with dying <100 times (but arbitrary training on video beforehand) seems like a nice candidate for another fire alarm (I wonder whether people would panic then, or whether that challenge is already AI-complete)
[5.7 months ago]
paradises, peacocks, peaches and palimpsests! Prussians, prowlers plus plumous penguins, priestess parks principles – priceless.
[5.8 months ago]
the existence of competence porn implies the existence of competence horniness
[6.0 months ago]
it is cold in the anthropic shadow.
[6.0 months ago]
a bunch of money in celestial futures: long on angels, but the bulk put into ein sof
[6.0 months ago]
my single-peaked ass
[6.1 months ago]
censors will censor complaints about being censored. therefore we should complain about censorship all the time – once those complaints disappear, we know things are going downhill.
[6.1 months ago]
“= I've been waiting for this for years. I don't want to join in what I read is supposed to be a highly competitive work environment. But I am impressed by the vision and potential of your approach.” Pretty much my reaction to new many EA stuff.
[6.1 months ago]
“Two men in mirror saws the edge of time For real or woe, we bury the grass and raise the” cryptic
[6.1 months ago]
“ah, but I use myself as a handle”
[6.1 months ago]
Hah, tweet-like-me.xyz with my old twitter handle is great. “Agreed. Should we therefore call for »epsilon umeshisms«?”
[6.2 months ago]
a model that gives a probability distribution over 12 orders of magnitude is good and brave and we need more of it.
[6.2 months ago]
@alip "Br em hlk rmnnd, Fkrdm, bkldryrum! Ben br bk ucuyum, Kvg verm hlkn, Bkldryrum te, Vrn benm frkm ! #sydbx" -> Aoe iiiu! Iyou eiau üuiu yäiiö. Äuuuůǒ ēēiyou #ïïïyae
[6.2 months ago]
i remember having quite strange & unusual qualia as a child (kind of hard to describe & varied, i remember 2 specifically (not just the old “having more imagination” or “being easier to frighten”, but on a similar level as déja vu)), but these have fad...
[6.2 months ago]
I'm not an intellectual, I'm just a guy who likes to f**k.
[6.2 months ago]
@(frogorbits.com) "@niplav interesting, it seems like you uploaded at least two weeks worth of posts sometime between yesterday and right now" -> Yes. I have changed my interaction with the internet a little bit – uploading/downloading in batches, not ...
[6.2 months ago]
gifted child? no. I am a childish gift
[6.2 months ago]
if there is a reachable maximal economic productivity of matterenergy/spacetime, then at some point economic progress will slow from exponential to cubic, since if one expands at a constant fraction of c, then the growth is proportional to the volume o...
[6.2 months ago]
theiropia, ouropia, heropia, hisopia
[6.2 months ago]
youropia
[6.3 months ago]
globe-decorated neo-neoliberals were furiously fruitlessly arguing with zero-summers on twitter, shaman bros were helping others get laid at parties funded by eternal-life cryptomoguls, infra-red haired identity-fractalites with two dozen Mercedes star...
[6.3 months ago]
identity politics defection from class warfare, maybe?
[6.4 months ago]
the inclusive genetic fitness party
[6.4 months ago]
trump turned out to not be such a big deal
[6.4 months ago]
gonna start crediting Carl Shulman for some of my ideas, seems trendy right now
[6.4 months ago]
Ah! “Logical Prior Probability” (Demski 2008). Of course
[6.4 months ago]
Niplav Yushtun, treasurer of the “Clone John von Neumann a Million Times” party
[6.4 months ago]
why no paradise engineering companies?
[6.4 months ago]
not reading
[6.4 months ago]
the singular of equilibria is equilibrium goddamnit!
[6.4 months ago]
impact measures depend on, rather than specify, inaction baselines
[6.5 months ago]
bullett
[6.5 months ago]
meditation techniques have inbuilt anti-abuse mechanism if they reduce agency and/or wanting
[6.6 months ago]
basin of corrigibility seems to be a reasonably large crux for prosaic alignment vs. miri debates
[6.6 months ago]
don't be middle class
[6.6 months ago]
[6.7 months ago]
nice bullet you got there. would be a shame if someone were to…… bite it
[6.7 months ago]
impac
[6.7 months ago]
metis, doxa, episteme, gnosis, techne – did I forget anything?
[6.7 months ago]
because prune needs to be more calibrated than babble
[6.7 months ago]
or: sneering=uncalibrated pruning
[6.7 months ago]
sneering is pruning without any metric of success
[6.7 months ago]
left not as good at babble as it used to be? more pruney, definitely
[6.8 months ago]
no canonical universal turing machine for solomonoff induction
[6.8 months ago]
………women
[6.8 months ago]
and, not to forget the obvious contender,
[6.8 months ago]
……girls
[6.8 months ago]
the burrito judge needs to be extremely powerful
[6.8 months ago]
formalise early, formalise often
[6.9 months ago]
ikto verso lapido
[6.9 months ago]
does the buddha have maze nature?
[6.9 months ago]
rational people can use very irrational people as babble generators in conversations, if the rational people are high prune (which they usually are).
[6.9 months ago]
qouta
[6.9 months ago]
not yet woke
[6.9 months ago]
cannibalistic respectability cascades in the water supply
[6.9 months ago]
memdemic
[6.9 months ago]
[7.0 months ago]
(mostly)
[7.0 months ago]
excess wealth spent on addiction or sex
[7.0 months ago]
approached a girl on saturday. the approach itself was nothing special, & i got rejected as expected. BUT I know understand what all those guys mean with femininity. this girl was dripping with it. she even gave me a compliment (“you have beautiful eye...
[7.0 months ago]
looking forward to the moment when i can first smack someone with my copy of Grand Futures because they're being an annoying pseudo-primitivist – the satisfying sound of a thousand pages of industrial strength paper against their weak biological skull
[7.0 months ago]
carteography
[7.0 months ago]
MFTOM
[7.0 months ago]
MFTOC
[7.0 months ago]
how come nobody takes the view seriously that coordination is eternally hard on all levels?
[7.0 months ago]
discourse between a man and his baaaaaa
[7.0 months ago]
at least we'll die a really cool death
[7.0 months ago]
what a strange optimizer it is
[7.0 months ago]
a date with a human model
[7.0 months ago]
todo: find some nice population axiologies where the bound is a parameter
[7.0 months ago]
makes me update towards the awfulness of ancestral environment
[7.0 months ago]
ancestral social environment probably most comparable with today's high school – everybody is very near each other a lot of the time, high amounts of common knowledge. differences: existence of age gaps in AE, more common tasks
[7.1 months ago]
gridworlds scenario with corrigible/noncorrigible subagents
[7.1 months ago]
“improve the system” is not an atomic action
[7.1 months ago]
https://www.lesswrong.com/s/7CdoznhJaLEKHwvJW/p/LfGzAduBWzY5gq6FE uses color really well in proofs. I will need to copy that.
[7.1 months ago]
forecasting questions can be divided into “we have an indicator, let the people forecast it” and “we have a question, let's find an indicator/indicators to concretise it”
[7.1 months ago]
Gotami said: “why is experience bad so often?” Upon hearing this, the nun got enlightened.
[7.2 months ago]
internal/external validity for attractiveness rankings
[7.2 months ago]
heiliges römisches rotes kreuz deutscher nation
[7.2 months ago]
even an ASI would defer to an even smarter version of itself. does this solve the problem of fully updated deference?
[7.2 months ago]
twitch, robot, twitch!
[7.3 months ago]
toe tall, I tear ian?
[7.3 months ago]
@iolfree "TEST hello twtxt xxx" -> :wave:
[7.3 months ago]
amazing they managed to get double-blind RCTs through asymmetric justice
[7.3 months ago]
christianity has been bad for sexuality, do other religions fare better? monotheistic ones don't seem to
[7.3 months ago]
tibeetan dreeem yooeeeeeeeee
[7.3 months ago]
guy who has always wanted to get into lucid dreaming
[7.3 months ago]
(yes, I know the thought experiment doesn't come from buridan)
[7.3 months ago]
buridan's ass: not sure which cheek to clap
[7.3 months ago]
“considered harmful” is not about harm: an opinionated review is all you need
[7.3 months ago]
if MDMA is libido-enhancing for women, and libido-diminishing for men, then that points at an important joint between male and female sexuality.
[7.3 months ago]
the gearless gear
[7.3 months ago]
I put on my best reference class tennis outfit
[7.3 months ago]
bell's theorem for the haecceity of qualia
[7.4 months ago]
yay ingroup
[7.4 months ago]
markov as inverse lindy
[7.4 months ago]
That's pretty Markov of you, my friend
[7.4 months ago]
DSOTM=Doesn't seem obvious to me.
[7.4 months ago]
philosophical bullets i'm not (yet) willing to bite: illusionism, accepting the repugnant conclusion.
[7.4 months ago]
“showing that you care” type signalling depends on the sunk cost fallacy; or, there was selection pressure against the sunk cost fallacy (modulo reputation effects)
[7.5 months ago]
Robin Hanson is pretty good evidence for the Orthogonality Hypothesis
[7.5 months ago]
how much would you X yourself?
[7.5 months ago]
or, worse, a minus
[7.5 months ago]
I dislike it when people use two en dashes for an em dash
[7.5 months ago]
wondering whether non-compartmentalization is a result of monotheism
[7.5 months ago]
boneless intellectuals
[7.6 months ago]
i want a smallpox vaccine
[7.6 months ago]
putting myself on a list by googling various bioinfohazards
[7.6 months ago]
6fe54e3e669470cf6b3b371149716d06704beda48a0ca553f45cf4a913edd180
[7.6 months ago]
“Attention is all you need”. Fucking prophetic
[7.6 months ago]
mesamodernism
[7.6 months ago]
bayesically,
[7.7 months ago]
PIPO
[7.8 months ago]
every man is a pick up artist. some even know it
[7.9 months ago]
a microwave without the urge to beep
[7.9 months ago]
it's just a set bro, it doesn't even contain anything
[7.9 months ago]
preposterous to assume that mathematical objects are real
[7.9 months ago]
*everbody* had a hard time in highschool :thinking_face:
[7.9 months ago]
middle manager of the local dharma
[8.0 months ago]
the good thing about this is - …………uhh…………………
[8.0 months ago]
muggee
[8.0 months ago]
“those motherfuckers actually read what I write. lmao” – Nick Land
[8.0 months ago]
it's important to remember: people rarely fundamentally change
[8.0 months ago]
alignment posting gf, capabilities posting gf
[8.0 months ago]
tonalli gf, teyolía bf
[8.1 months ago]
the success of discord and its consequences have been a disaster for online information flow
[8.1 months ago]
deontology more embedded than consequentialism?
[8.1 months ago]
guy who has had unkind thoughts about Eliezer Yudkowsky
[8.1 months ago]
guy who is building AGI on a discord server with his buddies instead of getting himself a girlfriend
[8.2 months ago]
marvel and be wary
[8.2 months ago]
someone (not me) should try to learn the 1k most common words in as many languages as possible
[8.2 months ago]
FreeBSDick
[8.2 months ago]
Venn diagrams work better on hyperbolic surfaces?
[8.2 months ago]
Let A be the set of AI alignment complete problems. What is the problem a /∈ A so that solving a maximally delays the development of AGI? What is a, weighted by how hard/desirable it is?
[8.3 months ago]
Yes, I can shoehorn interpreting extropianism into every poem by Schiller. This is going to be a regular thing.
[8.3 months ago]
This is obviously about O'Neill cylinders, empathetic pearcean valence, and the supercooperation cluster
[8.3 months ago]
Was den großen Ring bewohnet / huldige der Sympathie / zu den Sternen leitet sie / wo der Unbekannte thronet
[8.3 months ago]
@movq "Laphroaig. 🥃 🥴" <3
[8.4 months ago]
consequentialist buddhism