welcome to ghosts

In a strange position where I think there are massive, massive social harms to the proliferation of AI (misinfo, deepfakes, predatory 'companions', surveillance, labour rights) but also feel the mainstream arguments against it range from unconvincing (water) to actively harmful (copyright; "soul").

Like, I often hear "plagiarism machine that's killing the planet" but I'm very convinced that the copyright panic is actually harming artists far more than helping them. Strengthening Disney and Getty's IP rights in the name of "protecting artists" is a very bad thing, I think!

People often compare "AI bros" to "NFT bros", rightly so (often the same obnoxious people). Yet in the same breath, many will also essentially say right-click-saving an image is theft. I can't stand the spread of slop either, but I think that argument could set a dangerous precedent in other ways.

Or, you frequently hear "well, analytical AI is good, but generative AI is bad", but there isn't really a hard difference between the two. They often share identical underlying architecture. Like, transformers power both BERT and GPT. Vision Transformers handle tumour detection and image generation.

Moreover, tools classified as "analytical" are frequently used for facial recognition that automates surveillance & discrimination, and tools considered "generative" like diffusion models are also used in medical imaging for anomaly detection and classification. There isn't a hard "one good, one bad".

tl;dr on this one: the loudest AI criticisms are the ones that serve existing power structures, and yet its most serious harms are the ones that don't.

And it's a frustrating thing to see, because it's like, who benefits from the messy conversation around this stuff? The copyright panic benefits Disney/Getty/Adobe/etc. The analytical vs generative framing benefits companies selling surveillance tools. Presenting environmental stats taken wildly out of context benefits people who want to dismiss AI criticism entirely by making critics look innumerate. Meanwhile, the actual harms—like predatory companion apps targeting lonely teenagers, facial recognition enabling police harassment of Black neighbourhoods, algorithmic benefit denial systems destroying families, labour displacement without social insurance—don't have wealthy institutional advocates the way "protect our IP" does.

And then, I don't know, the criticism also becomes weird and individualistic. Like I have a dear friend who is a disabled artist who writes her own code for image gen and gets torn apart for it on the regular with straight up ableist disability porn ("X person drew with their feet, why can't you?").

Like, to be clear I think these companies are by and large straight up evil. Sam Altman releasing 4o is like. Practically criminal. Musk actively removing safety guardrails as a selling point, deploying it on a platform that's already a misinformation vector to generate racism and child exploitation.

But I just worry the current way these things are spoken about ends up actually obfuscating those harms. I might roll my eyes at some rando using an LLM to write their emails, but like I don't buy that they're doing something any more evil than taking a flight or eating meat or leaving the lights on.

And this kind of individualized critique is just the "ecological footprint" thing again, literally invented by BP's ad agency in the early 2000s to shift responsibility from fossil fuel corporations to individual consumers. The impact matters, but I wish it wasn't being discussed on Shell's terms.

AND I think that framing prevents good regulation. If your position is 'this shouldn't exist at all and I won't tolerate anything less than the technology not existing,' it's harder to advocate for 'here are the specific features and practices that should be required/prohibited.'

I strongly look forward to there being less of this shit everywhere. Every product suddenly has an AI feature nobody wants. LinkedIn AI-generated posts. Google AI overview and its wrong answers. Facebook flooded with Jesus made of shrimp. That viral UberEats fake Reddit post from the other day!!!

There's so much marketing hype; all the hype and some of the hate are often doing the same work of avoiding talking about its biggest dangers. I'm sick of AI bullshit everywhere, afraid of its worst applications, and also disillusioned with the way it's being talked about in many of my circles.