dvorak edm remix which is actually good
tolerating what we can
taking the pledge to become 10% less misogynistic
now is a good time to sit back and read
topics i will never research: how the prussian military inventing boardgames with dice to train military strategists influenced modern eurogames and not amerigames
It seems quite unlikely to me that humans will cease to care about wealth after the singularity (15% maybe?), or that markets will cease to exist after the singularity (30% or sth). so the best investment strategy now for post-singularity scenarios is to invest broadly in the economy. not sure about divesting from AGI companies (bc they accelerate danger) or AI hardware companies (ditto)—they’re going to especially valuable post-singularity.
L’an 2100 ne passera pas
from now on, when talking about fractions, i will only use the terms “upper number” and “lower number”. those both have fewer syllables AND they’re vastly easier to understand. Thank you for coming to my TED talk.
libsucker. rolls of the tongue
all the cool kids? in the bay. me? chillin’
HAHAHAHAHAA IT’S FEBRUARY 2020 ALL OVER AGAIN HAHAHAHA
warm & wet game theory
being cone-theoryd
solution to the fermi paradox: ~all species don’t have nerds bc nerds are locally maladaptive, get stuck at the social reality bullshit level, humans are a fluke in the regard (kindness to kin, but for…, uhh, nerds?)
i think in practice people were most convinced by timelines, if not arguments, then observations abt progress.
using subscripts for probability assignments?
malign prior as an idealized case of meta-optimization
lw events should give one the ability to report probabilities
did women really faint that often in the past? if yes, what changed so they don’t anymore?
the new chinchilla scaling laws were not what I expected
whenever someone denies something, nod my head sagely and say “glomarizing nicely, I see.”
shoutout to the woman that broke my heart so that I now read papers titled stuff like “Multiverse-wide Cooperation via Correlated Decision Making”
HOW DO THOSE PEOPLE READ SO MUCH
that probably relies on a bunch of interpretability work done beforehand
todo: find a way to track ontology shifts in neural networks
me reading planecrash is unstable bc while reading it I become so enthusiastic about reading textbooks that I go off and do that instead
just thought to myself “hopefully a bigger pandemic hits, that sounds like it’ll delay ai capabilities progres”, which, no,,,
a rationalists lament: on the EA community
what are people’s attitudes towards others voting selfishly vs. not voting at all?
myopic consequentialism might often outperform heuristics, don’t know about full consequentialism in bounded agents tho
biology:philosophy of mind::ethology:meditation
researchers drunkenly explaining 5-10 to each other
surveys probably get more results if people get their some summary statistic at the end—a way to increase participation rate?
find the next number in the sequence 0,2,2,46,3640
daily reminder that most politicians wouldn’t be able to explain the concept of expected value to you if you asked them
I’m a midwit with high need for cognition
reading reddit thread about advice for women fighting men rn, and one of the advices is to yank an eye out, not considering that eyes are pretty hard to yank out
commitment races in rape accusation
rip schelling.pt
stiglered again
pet peeve: scales that have a neutral or default or average point, but which are nearly or completely on the positive. why are people asked to rate happiness from 0 to 10, not -10 to 10, where 0 is the neutral state? why are movies rated from 0 to five stars?
the right level for solving the hard problem of consciousness is within existing science/within philosophy/within meta- or pre-philosophy/needs a fully new paradigm of thought
transhumanism is simplified humanism.
ideologies are just compressed & distributed vibes, shards of shared stances.
fodgehox
“im not a furry, though, i just listen to the music”
Einstürzende Neugeburten
good hobbyist internet science writers & scientists are afflicted by a peculiar disease: they produce some (or a lot of) good output, and then someone gives them enough money to start an organization, and they never write anything interesting again.
ganrsl
i’m now pretty sure the urbit whitepaper is impossible to parody.
oh yeah I love software that just dumps giant amounts of data into my home directory without giving me any option of changing where it dumps the data
GPT-Alyx (Codebase: 96% Idris)
i find the removal of the “master/slave” thing in computer protocols to be kinkshaming
the EA community is grabbing existential risk by the black balls
maybe for most people social media isn’t an epistemic hazard, but “just” an attention hazard?
there’s a lot of alpha left in high-weirdness mental wellbeing interventions
lojban rap, microtonal giantstepscore, taxxonpop
if anything indicates anything, the french are now to be feared.
pandas makes my RAM usage jump like a cardiogram
if you want to minimize the amount of words in your library, every function should have as many parameters as there are functions in total.
plasmacrane
I expect the future to be more, not less, absurd than ~all scifi. way more—there’s no scifi this abstract & absurd, and noone would buy it (& also bc of Vinge’s law)
dath ilan has screened off its past just es ey wants to screen off his past before ~2003
numpy.unique needs the type to be comparable by < and >, not just ==. that seems like overkill.
diversity inequity exclusion
following legislation from the ELSC (european long site commission), I now have to adopt an ocelot
apparently we’re going to build AGI, whatever that is
evopsych is confused by persistent conflict between mothers-in-law and spouses—don’t both want to maximize the number of children? i don’t see any conflict arising for evolutionary reasons
it’s crazy how we basically haven’t tested nuclear weapons in ~30 years
how much did large volcano eruptions in the past change agricultural output?
you know that you’ve made it when you’re using fermi estimates to debug your rl agent.
rationality is having strong opinions about things nobody has ever thought about before.
lipogrammatic pangram
found spicy khmer waltz vexed quagga jacob (Mr. Jock, TV quiz PhD, bags few lynx)
extra spicy bunns
naxus
foqed
jokez
ezafu
jarns
janus
mozed
bouge
foofe
boxem
neither is warez
qream
spicy wheat found
oofed not in the absurdle word list, even though it’s the reaction i have the most when playing it
“if enough of us listen to prophets of doom, it becomes self-fulfilling” — Leo Szilard
pressman being another taxxon fan is a delightful finding
if your eag profile doesn’t say ‘neartermism’, you’ll automatically be drafted into the deepmind potential hires mailing list
“The problem with Marcus’ argument is that the only alternative to statistical AI is spiritual AI“ no what the fuck why would you say this did you even read the sequences i don’t even know where to start with that
holy fuck were ACX comments always this bad? most of them are fucking atrocious. like, barely tolerable.
at least i’m up to a level where i’m reading a comment on an acx post and someone uses “consciousness” and i’m hella out there real quick
to what extent did AlphaFold2 solve protein folding? there should be a writeup of how much & what needs to be done for it to be completely solved.
wait, is CIRL incorrigible for the same reason that utility-maximizers don’t wirehead? https://niplav.github.io/notes.html#A-Short-Example-For-Why-CIRL-Is-Incorrigible
solarpunk slightly hindered by the fact that for solar panels and cutesy robots you need industrial-strength chipfabs that are so advanced we have like 6 of them in the entire world