Show newer

me: man i wish MIRI shared their models publicly more I love reading what EY has to think

KOROK the GOD of shit and piss and slime and cum and rot ripping bacwards the flayed ape's hand: Granted.

then i can idfk get the weights to glow on parts of the net that are an optimiser or something so that OpenMind can see what they're doing and decide not to so UNFAIR can deploy their thing 6 months later

Show thread

alright getting up to date with pytorch. last time i did a neural network it was a lovely multi layer perceptron neural network in keras but that barely cuts it for the kaggle tutorials nowadays

In Tetratopia the paper credit representations have “in Science we trust” written on them

won't the nukes be cold and lonely, trapped in their silos?

the sluggish and dimwitted Sioned, widely whispered to be touched in the mind, was cruelly nicknamed "the Wise" struggling with her childhood topology lessons

kiuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuyeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeedqejjjjjjjjjjjjjjjjjjjjjjjjjjjjj jujjjooooooooooooooooooeooooooooooooooooooooooooeoqoqqqqqqqqqqqqqq;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

Show thread

=/=\/\\\\\\\\\\\\\\\\\\\\=\=====-=------------------lrrr0rcccccccc-c9tttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttttwwwwwwwwwwwwwwwV wwwwwwwwwwww wmmmmmmmmmmmmmmmmtmmbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbhhhhhhhhhhhhhhhhhbhggggggggs8gddddddgddvdfxxxxxxxxidiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiwdiiikkkkk kiikk

there should be a website grok.io that explains ML papers in simple ways so that I can be really annoyed at the name (for the same way some people dislike "moist")

maybe low-effort fun even. it's not social because social media by default doesn't count as social, unless it's a DM or a proper 1:1 conversation. the archetypal example of doom is twitter, but you guys are more endorsed than twitter

Show thread

timesheet codes (posting this is keyed as Doom but it's part of the Life code of setting up the timesheet so I'll permit it)

Show thread

Garry Kasparov can't exist. Whenever I ask people what chess moves he would make, they just smooth it over with a vague and magical "one that's good"; the only time someone deigned to give an actual theory of how he'd win, it was a long conjunctive string of moves and coutermoves that I could have easily completely invalidated just by making a different move earlier in the game. Clearly it's just the delusion of someone who's watched too much Queen's Gambit.

we need a new AI paradigm based around trapping hostile ghosts because then finally apparently everyone's intuitions will be correct

Show older
Mastodon

a Schelling point for those who seek one