nick m
@diskontinuity
Still sailing east.
cant we just train a massive autoregressive model on like literally all of someones sensorimotor IO and according to (the simulator hypothesis? idk if this has a name yet) get an epicycle model of their mind
the church-turing thesis is no more. the church posting shall continue on this account. the turing posting will continue on the other account [iykyk].
made a priv so i can despair + fantasy post in there instead of shitting up the tl on here. only pure signal from here on out [i hope].
what is with the whole "X doesn't matter, that's why instead of calling group A 'un-X-ers', we're gonna call them 'aktchually-X-if-you-look-at-it-this-way-ers'"? it seems like a self-humiliation ritual. except its majorities doing it to minorities so its just humiliation lol.
I think liberals might underestimate how much of their project functions and is agreed to because of the sense that white people are a stable majority with significantly more power than minority groups. I suspect that this kind of language emphasizing the opposite arrangement is…
the way qts are formatted sets up a really ugly accumulation of attention for completionists (flobn)
I think one has to understand re. the googolshrimp problem that human reasoning is really fundamentally about bounded rationality, and as soon as you cross those bounds you have to give up a bit of your humanity in one way or another.
a simple yet common sign of bad design is when effort scales inversely with task complexity. I ran into this when I was sorting my files on Mac. I have a lot of screenshots on my desktop and I want to go through them one by one quickly, and sort them into folders manually. ...…
im realizing its so important to keep choosing to be strong in small ways. i didnt really understand this until i finally saw how so much of my problems were downstream of the precise opposite. and if choosing to be weak in little ways can add up to despair, misery, then maybe...
Always do X. Wait, don't ALWAYS do X, there are obvious edge cases. But if you try to distinguish those edge cases on the fly, you end up stumbling into worst-cases; save yourself. So you just need to move your heuristics accordingly. But you can't trust yourself unless you fully…
Ok, q: there are Turing machines whose halting behavior is independent of ZFC. But we know in some sense these machines "can't really halt"; else there would be a finite proof of halting: just run the TM for that many steps. How much categoricity can we force out of ZFC this way?
Don't you get a sense of longing seeing something grand humans built? Not awe, latent frustration. Not just bridges and cathedrals; complex emotions, personalities, _language_. This is what it feels like to look at a superintelligence. Something orderly, yet beyond you — chaotic.
Gemini: "yes, this is all very beautiful, but do try to be the one holding the knife, not the one bleeding out."
