Post

All of the LLMs stole from me. Most of the big "A.i." billionaires are horrifying right wingers. Consumer-facing generative "A.i." "writing" and "art" tools deskill people, cheat kids out of real education, spread falsehoods and disinfo, and violate privacy. So no, I ain't using that shit.

I've always said there are legit uses for machine learning in science. That's great! Go for it! But I will likely never be interested in the consumer-facing generative and LLM bullshit that's linked with fraud, cheating, theft, right-wing ideology, deskilling, and even murder & suicide.

158 reposts961 likes

Also: the "A.i." billionaires are exploiters and liars and extractionists. Most of them support right wing fascists. They literally cannot be trusted about anything. So I'm not interested in participating in anything that furthers their interests or gives them more money.

I don't subscribe to any Substacks because that company has platformed and promoted Nazis. I switched away from Shopify because they platform horrible people. I deleted X because that fucker is one of the worst people on the planet. Why the hell would I ever use an OpenAI product?

Train up your own LLM and run it locally as a GFY message to Sam Altman, Zuck, Larry Ellison and all the tech oligarchs. Well except for those who sell the hardware. And the energy companies. And Internet Providers. And ..

Yeah, yeah, there's no ethical consumption under capitalism. And yes, I still use some services provided by corporations run by monsters. But I'm doing my gat damnedest not to buy any NEW shit from obvious monsters.

No objection to AI use for advanced pattern recognition, but genAI for text & images is a dead end, and unfortunately that’s what the entire industry is hinged on. I think a lot of the animosity people have towards AI is based on the intentional, dangerous misuse being encouraged by profiteers.

There’s a place for the technologies these companies have built, but not at the scale they’re hoping for, and not with the sycophantic chatbot interfaces they’ve used to humanize it.

It should also be noted that the ML technologies that are "useful to science" existed at least as far back as the 1990s. Those models are focused, dedicated tools for investigating specific problems. I remember running protein folding programs to "help science" as a '90s kid in my bedroom.

LLMs are at BEST generalized tools that can perform a poor imitation of these "useful science tools". The majority of their training data doesn't pertain to your problem, and actively makes the tool worse at investigating your problem, you're always better off making a dedicated tool.

The "it's good for science" argument is basically stolen valor. It's marketing. They conflate their crappy plagarism engines with genuine useful scientific inquiry, because it convinces uninformed people that what they're doing is important and useful and deserving of support.

Meanwhile, they've gobbled up TRILLIONS of dollars of investment, while vaguely promising to cure cancer or solve climate change - knowing full well their technology can't and won't do that. Imagine if we had put a trillion dollars each into renewables infrastructure and medical research instead?