slivers of a million lives

mind over matter. against sentimentality. for snobbery. engineer fake fusion phycisist.

[archacovercosine]

waren wir nicht einmal hier in diesem Wunderland?

gingen wir nicht einmal staundend hier entlang, an uns selbst vorbei?

[citationzero]

"For simple installation use Conda" <- fuck offffff

[archacovercosine]

in a just world tastier food would satiate you for longer than bland gruel

[archacovercosine]

there are still people around using "my sweet summer child" totally unironically, what a crazy world

[oxeanic]

catfishing.net

718 - 7/10

🐈🐈🐟🐟🐈 🐈🐈🐈🐈🐟

[archacovercosine]

catfishing.net

#718 - 5/10

🐟🐈🐟🐈🐟

🐈🐈🐟🐈🐟

[papergemstone]

The Amazing Digital Purpose of an Amazing Digital System is what it Amazing Digital Does

[archacovercosine]

for mostly understandable reasons the balkans and late-ottoman empire constantinople appear to be the favorite inspiration for writers of the "what if there was a weird city" subgenre that may or may not coherently exist.

[archacovercosine]

during the largest part of the 20th century, the dutch protestants, catholics, social democrats, and liberals,all had their own political parties, broadcasting organisations, newspapers, schools, trade unions, sport clubs, universities, etc.

this system eventually collapsed under various political and societal changes, forcing for example the protestant and catholic parties close together to protect their common interests such as religious schools, but the mechanisms for this system are largely still in place.

[xylophonetangerine]

“Padel” is like squash except worse and even more fascistic.

[archacovercosine]

pickleball for investment bankers

[archacovercosine]

the thing with ai safety is the conviction that a sufficiently high intelligence essentially gives you mind control powers. which i don't think is true.

[kaiasky]

idk, i don't think this has been the serious model for a while now. or like the classic setting for invoking ~braingenius mind control~ is "the AI is trapped in a box and wants out", as like, an argument for why boxes are a bad solution. (and at the end of a list of other ways to get yourself out of a box. "even if we had an unhackable box with armed guards and the ai's words were printed on paper and we all promised to never run any code it wrote... it could still just try talking to us.")

(this looks rather different than the current state of "the ai writes the code for the box and also much of major web infrastructure"!!)

but like e.g. a major part of contemporary ai safety is "the ai could probably make turbo-smallpox and that'd be bad". The ais don't want to make smallpox, but could be trained to want to if somebody had their weights. insofar as boxes are involved it's. "the ais and the humans both agree that the weights should stay in the box"

idk. "turbogeniuses would necessarily have infinite rizz" doesn't seem like an essential claim of AI safety as it exists in 2026. which is not to say that it (and broader intelligence-dicksucking) isn't a common part of the ai/rat/bay area memetosphere ofc.

[archacovercosine]

yeah there's definitely a distinction between safety wrt to an autonomous superintelligence and safety wrt to humans steering the AI. I was under the impression that even though most concrete work is in on second case, the people working on it also all believe in the first case, with the way the look at these problems also influenced by that belief. But I don't live in the bay area so it might not be true.

[no-sharks-on-the-sun]

i think the threat models for autonomy and human steering are basically identical at this point and (as kaia pointed out) neither involve persuasion

[archacovercosine]

we might see the inverse: a human operator persuading (tricking?) the model to step outside of its boundaries

[archacovercosine]

the thing with ai safety is the conviction that a sufficiently high intelligence essentially gives you mind control powers. which i don't think is true.

[kaiasky]

idk, i don't think this has been the serious model for a while now. or like the classic setting for invoking ~braingenius mind control~ is "the AI is trapped in a box and wants out", as like, an argument for why boxes are a bad solution. (and at the end of a list of other ways to get yourself out of a box. "even if we had an unhackable box with armed guards and the ai's words were printed on paper and we all promised to never run any code it wrote... it could still just try talking to us.")

(this looks rather different than the current state of "the ai writes the code for the box and also much of major web infrastructure"!!)

but like e.g. a major part of contemporary ai safety is "the ai could probably make turbo-smallpox and that'd be bad". The ais don't want to make smallpox, but could be trained to want to if somebody had their weights. insofar as boxes are involved it's. "the ais and the humans both agree that the weights should stay in the box"

idk. "turbogeniuses would necessarily have infinite rizz" doesn't seem like an essential claim of AI safety as it exists in 2026. which is not to say that it (and broader intelligence-dicksucking) isn't a common part of the ai/rat/bay area memetosphere ofc.

[archacovercosine]

yeah there's definitely a distinction between safety wrt to an autonomous superintelligence and safety wrt to humans steering the AI. I was under the impression that even though most concrete work is in on second case, the people working on it also all believe in the first case, with the way the look at these problems also influenced by that belief. But I don't live in the bay area so it might not be true.

[archacovercosine]

the thing with ai safety is the conviction that a sufficiently high intelligence essentially gives you mind control powers. which i don't think is true.

[archacovercosine]

i guess there is safety in the sense of guarding individuals and society from other humans giving the ai instructions, amd safety in the sense of guarding humanity from a independent superintelligence.

idk about the second one but I think it would be bad if we were too aggressive on the first front

[archacovercosine]

the thing with ai safety is the conviction that a sufficiently high intelligence essentially gives you mind control powers. which i don't think is true.

[archacovercosine]

well i mean human (or not) consciousness and what's going on in the brain is one of the very few places in the collective knowledgespace where god is not utterly dead yet, so of course people of a certain bent flock to ideas about it.

[archacovercosine]

and what's always really fun about this is that even the question if we will eventually understand it is part of the discourse.

[archacovercosine]

well i mean human (or not) consciousness and what's going on in the brain is one of the very few places in the collective knowledgespace where god is not utterly dead yet, so of course people of a certain bent flock to ideas about it.

[archacovercosine]

kind of curious how well fable or w/e does at generating spice netlists for various things. given that you run into tricky components like specific op amps pretty quickly.

will test if i ever acquire significant frontier access i guess

[raginrayguns]

the worst case scenario i guess is that carbon dioxide reduction becomes so cheap and easy that we actually have too little carbon dioxide for plant growth, but also we still have global warming because of byproducts like methane

[archacovercosine]

afaik methane decomposes into co2 in the atmosphere over time so that doesn't seem that critical? there is also a water vapor contribution to radiative forcing, not sure how that is increased or decreased.