Post

when you ask 85% of people on here why they hate AI or LLMs, they'll give you thought-terminating cliches or factually incorrect information it's a damn shame, because it weakens their positions when the correct and reasoned 15% is MORE THAN ENOUGH to tear the entire industry down

but no, it's "it's just autocorrect" and "it can't solve novel problems" and "it's just regurgitating what's in its training set" and "it's slurping up lakes" none of those are true. go after power use. go after everything capitalists are using it for. those alone easily show we aren't ready for it

even with the state of models today, they could do wild things if the primary reason they're being pushed by the sociopaths in charge wasn't "slash labor cost, destroy worker power, and replace people in automatable creative positions wherever possible" and if they were efficient.

they aren't power-efficient. just that disqualifies them from being something that's ethical to develop at the scale capitalism demands when we are deep in a climate emergency

like, christ, look at it: we're doing large-scale enough analysis on text to create an assemblage of reverse-engineered mathematics that behaves in ways complex enough to *solve arbitrary problems.* that's fucking WIZARDRY. and what's it being used for? evil, of course. we live under capitalism

principled (as in, deriving from principles) politics are hard to come by. :/ having an ironclad "technology should serve the people first" stance requires the acknowledgement of the global divide of labor which trips the anticommunist brainwashing in people.

that anthropic paper about "functional emotions" goes to great length to point out they're not saying their model has emotions. it's about internal representations of an LLM's state and how those affect its behavior. it's describing interesting and potentially useful math.

are you gonna refuse to read it and enter the shitty Wired headline in your memory instead, or are you gonna put any effort into learning how the things you hate work so you can actually understand what you're talking about when you say they should be abolished?

"spouting incorrect information about something you hate" and "saying 'well any reason to hate it is good, so I'm not gonna bother to be correct'" are two of the online behaviors that drive me up the wall most and nothing seems to evoke them more reliably than AI think about what you're thinking!!!