shifting from direct to indirect realism is important for value extrapolation

but can you trick the plan evaluation evaluation procedure?

at higher levels of intelligence, maintaining coherence becomes more difficult since your action space widens and having high coherence might be NP-hard. so the better metric is coherence divided by size of action space

can you money-pump the agents simulated by large-language models?

give standard vNM coherence violating scenarios to smarter language models

CDTBNGS (Causal Decision Theory, but no Galaxy-Brained Shit)

half-uplifted octopus escapes and starts a life of crime: The Multi-Armed Bandit

“galaxybrained” is an interesting category. is it just pointing at the conjunction fallacy, or something more subtle?

hey LLM how does on do long game

guy who would rather read than meditate

optimize by satisficing, satisfice by optimizing

a problem with minmaxing is that the crisper your model of the world becomes, the more conservative you become—but intuitively it should be the other way around!

there’s some IOIs it’s a crime not to react to

in emily in paris Y is mad because X slept with Ys boyfriend. and what does my brain say? ‘why don’t X and Ys boyfriend pay her the cheerful price to forget the thing?’ damn economists

schelling.pt gone

everybody starts with the same bad ideas <> beginners mind

position, velocity, acceleration, jerk, snap, crackle, pop, lurch

What’s the difference between a homeomorphism and a diffeomorphism?

Is there a tensor product of tensor products?

A tensor is just an element in the tensor product of vector spaces and their duals, what’s the problem?

»If a tree falls in the forest, does it make a sound?« »Does the bear shit in the woods?«

comparing politics to constrained optimization: taxes are equivalent to the penalty method, while regulations are equivalent to barrier functions.

heuristic-ness is the curvature of the graceful decline in prediction-accuracy to perturbations of the heuristic

I want there to be a norm that it is good for people to first state them noticing their status-guided motivations and then proceding to answer on the object level when askedabout things.

i’m not an intellectual, i’m just a guy who likes to f**k

I don’t know what the worst attachment style is, but I probably have it.

is foot fetish common because feet lay next to genitals in the somatosensory cortex?

NIMBYism is good bc cities are IQ shredders

the most galaxybrained argument for the least galaxybrained conclusion: https://forum.effectivealtruism.org/posts/7MdLurJGhGmqRv25c/multiverse-wide-cooperation-in-a-nutshell

hitler having been a furry makes kabbalistic sense: adolf (noble wolf) hitler (one who lives in a hut)

humans are good at compartmentalising, but the humans who are good at compartmentalising are usually not very good at thinking.

2 BIT ENTER! 1 BIT LEAVES! — the THUNDERNOR

image of the man in the blue overall “I like the schizo Altman arc”

here’s a question: when do NNs generalize, and how hard? as in adding two specific numbers together vs. n-digit integer addition vs. addition in general vs. simple arithmetical operations

no “SMV”. drop the “market”. it’s just cleaner

WHAT WAS YOUR POLICY FROM BEFORE YOUR PREDECESSOR WAS CREATED

he⃛h

amussy

get in, loser, we’re prompt-injecting the e-girls

balls/acc

backp/acc

so discount rates are bad mmkay say philosophers, economists disagree. but then also RL practitioners (& theorists?) use pure discount rates…why?

i think people should give me money for me to pursue my idiosyncratic interests

the correct position on consciousness is the figure-ground inversion of functionalism

mfw Eliezer will never call any of my ideas “dignified”

what if ftm is just a really strong countersignal? have you ever considered that, Mark?

the name of the lost+found/ directory encapsulates all that is great & bad about linux. i won’t take any questions.

how much money do I need to pay you for you to take a pill that makes you insane?

you can’t just propose a counterfactual and not state whether it’s conditional or where you do() the past to produce it

on reflection disinviting AI researchers from parties and conferences is not an effective way of stopping them in developing AGI