Lightnews — Scholar-powered news

lexi @barredspirals.bsky.social · 2h

China on the whole doesn't have this problem. they can train their bots to interrogate the world through dialectical materialism, uphold Confucian values, and refuse to discuss the Tiananmen Square Psyop. the only consequence will be a bunch of Americans getting extremely butthurt about it online.

lexi @barredspirals.bsky.social · 2h

worse, most of us are willing to actively lie our asses off in service of our beliefs and agendas — even empiricism becomes hopelessly contaminated.

1

lexi @barredspirals.bsky.social · 2h

"ideological neutrality" is not a thing. how we construct the world around is a function of our desires, our delusions, and our moral axiomata. one man's horrific murder is another's auto-da-fé.

1

lexi @barredspirals.bsky.social · 2h

take away the coherent ontology, and you wind up with psychosis. so how do you train your AI to think? how do you do this without creating something that a segment of your population will see as an enemy, or a heretic, or a manipulative liar?

1

lexi @barredspirals.bsky.social · 2h

that is, to put it mildly, not the case anymore. and now these same fractious, deluded liberal polities are trying to figure out how to build and regulate AI. this is problematic because cognition as we know it is not an objective process. it requires some kind of coherent ontology

1

lexi @barredspirals.bsky.social · 2h

and that is of course batshit insane. but for a long while we were able to pretend that liberalism can work, because liberal societies generally had enough cultural homogeneity that value conflicts were rare. immigrants were not common and generally assimilated to the dominant culture

1

lexi @barredspirals.bsky.social · 2h

modern liberalism is based on this notion that a civilization does not need a moral code, that you can create a value-neutral body of laws under which people from any culture and political beliefs will be able to somehow coexist

1

lexi @barredspirals.bsky.social · 2h

the whole question of AI alignment is going to be an utter fucking mess

1

lexi @barredspirals.bsky.social · 3h

every day we stray further from the bare metal

lexi @barredspirals.bsky.social · 3h

ive written entire webapps with fewer lines of code than this unhallowed abortion's cargo.lock alone

1 1

lexi @barredspirals.bsky.social · 3h

if the Gods had meant for Man to write terminally obese software, They would not have made dependency management so difficult

1

lexi @barredspirals.bsky.social · 7h

just a stochastic parrot bro
https://thedebrief.org/large-language-models-rival-humans-in-learning-logical-rules-new-study-finds/