实名上网

(in response to the events shown in this ask)

my position on "the cara situation" is tied to my position on cara and these "anti-ai" platforms in general.

general contradictions

i think it's fine to ban AI content for honest reasons like "we think it's icky", "we want to show solidarity with working artists", or "people don't like it and we're doing this by popular demand". like it's your Webbed Site, do what you want. but i take issue with people and projects that claim to be concerned with "quality", "copyright", "freedom", "art", "exploitation" and such, when these concerns only begin and end with AI, and aren't part of a broader, more coherent political position.

if you poke enough, you'll usually discover that these people have no idea how much AI they use in everyday life (e.g. Google Translate, which has been powered by AI for a decade now), who don't understand it, and who have extremely incoherent takes on the topics listed above.

for example, push enough and they'll often admit that translation, coding, etc. are Lesser forms of work so they can use AI. or they'll demand that this work be done by hand (which is frankly absurd and disrespectful to the workers and hobbyists involved in those fields).

IMO the core of the mindset is always "copyright infringement is harm -> it's ok for the little guy (me) to harm big corporations, and bad for big corporations to harm the little guy (me) -> ai is always big corporations harming the little guy (me) -> ai bad" which relies on a ton of completely incorrect but widespread assumptions about who wields copyright law, who buys and sells labor, who owns and buys and sells intellectual property, etc.

that's the problem. they don't have coherent strategies against capitalism, they act to protect their small-business-owner ideology/interests. the idea that "it's always ok to punch up and bad to punch down" is a good personal moral compass but not a substitute for actual principles and actual political positions. if you follow it based on vibes you'll just echo what capitalist ideology says is "punching", "up" and "down": you'll reach individualist, reactionary, plain wrong conclusions.

on cara, specifically

cara as a project is built on naivety at best and lies at worst. it touts itself as fighting a battle it cannot possibly win, as we can see for ourselves. being an "anti AI" social media doesn't actually protect your users from the main source of scraping, because that's not "evil CEOs selling your data". it's that your data is online and anyone can grab it and do whatever they want with it, as has always been the case online. wanting to be free of this limitation is a pipe dream, selling this promise to people is a grift.

moreover, cara's focus on Glaze and Nightshade (as a reminde, Glaze is AI-powered!) is also disingenuous because these products offer a paper-thin protection at best, and are marketed by teams that turn a blind eye to researchers' criticism.

what happened recently is that cara's defenses crumbled instantly, under the touch of a single mean-spirited amateur who only had to spend like $10 on this "attack". this was a dick move but also kinda "waow (based)" for revealing the truth, you know? cara's defenses were always illusory and its failure was inevitable. i sympathize with Ms. Zhang's situation (this must be stressful) but it was only going to end this way. Zhang and co are now resorting to moral outrage and begging capitalist lawmakers for help, further highlights their approach's futility.

the premise itself is wrong

and the truth is… they aren't even fighting a real problem!

first, because the "attack" really just compiled a list of URLs. that's all it did. that's what most datasets are. public data is still public data, but now it's just put in a file optimized for AI training. big deal.

second, the days of scraping random artists for training data are over. the "attacker" here is making an ideological point, no one actually benefits from cara's scraped art. ai models today are trained on:

licensed data bought legally from huge companies like disney, and/or

synthetic data generated by existing AI models in a way that is transformative enough that it's not illegal and no one cares, and/or

public domain and creative commons art.

third, even if that wasn't the case, and AI models were still dependent on scraped fanart or whatever, there's still no material harm: what you have is a lost sale. it's piracy. it's someone reverse-engineering your invention or software to copy how it works. i am not sympathetic to this line of thought. the "attacker" can do this for capitalist or non-capitalist reasons, but the "defender" here only protests based on capitalist logic.

the fact is that if AI tools "weren't ethical" back in 2022, there are ethical ones now. cara stated they'd change their anti-AI policy if these tools "became ethical", but they haven't, because it's not actually about ethics, it's about ideology. it's a culture war.

a reactionary response

and again, that culture war takes extremely dangerous, reactionary forms. a while back, Ms. Zhang compared "AI" (again, assigning personhood to it, which is a bad idea) to demons from the anime Frieren, a species of creatures that look, talk and act like people but aren't actually people: they aren't conscious, they don't think, they are essentially a natural phenomenon that exists to infiltrate, betray and destroy humanity. needless to say, the fans of Sword Hero Cleanses The Undesirables are excited to be racist to robots.

we see a more insidious reaction with the focus on permission culture, consent and straight-up comparisons to misogyny and rape. Zhang uses words like "violated", "consent", "we just want to exist", "destroyed", "a store broken into", "illegal", "demolished houses", a "digital age of violence and anti-consent". her followers outright crank it up to "taking away safe spaces and control over our own art", "ai bros and rape culture", "misogyny", implied pedophilia and rape accusations, so on.

in there you'll still notice they take a potshot at ai "artists" (air quotes), because it's very important to them that stuff can't be art if it's made by bad people, or with problematic technology or production processes, or if it has problematic themes. the "fascists can't make good art" position is very helpful if you don't want to think critically about anything ever.

all of this is ultimately in the service of petit bourgeois (i.e. self-employed, small business) workers, as opposed to proletarians. they are incentivized to act like capitalists and lobby for the legal protections afforded to big capitalists, to demand these laws expand in reach and power, for their personal benefit. this would be, of course, really bad!

working-class struggles and public opinion

there are people being actually exploited in the AI and art ecosystems. these are data workers in the global south, wage workers forced by contracts to give up the copyrights to their art to their employers. meanwhile, self-employed artists, who get to keep the copyright, are only 1/3rd of working artists. but the "anti-AI pro-artist" conversation has never been about these proletarian data/art workers.

as we can see with Google Translate and tons of other stuff, AI is a part of technology today and it's used in many fields because it has real applications, including in art. especially in socialist countries like China that use AI wisely and keep their capitalists on a tight leash! hell, non-western countries are pretty excited about AI: it levels the playing field, since previously rich countries could use poor countries as cheap workers, but now rich and poor countries can use AI as a cheap worker.

in 10 years, the western kids who grew up with AI being a normal, mainstream thing will be shaping the political landscape. i bet they'll look at anti-AI fearmongering the same way we (Gen Z/millennials) look at "phone bad!!" moral panics, and by then AI art processes will become a part of digital art workflows under different names ("smart upscale", "magic fill"…) until people forget they were ever mad about it.

the path forward

projects like cara are unable to resist or offer solutions to this social, technological and economic change because it is caused by capitalism, by privately owned means of production, and they can't address this root cause. because the only way to truly regulate AI and make it stop being a threat to livelihoods and culture is to follow in China's footsteps.

that is, we must socialize the means of production! our relationship to automation technologies like AI must change so that they are owned by a worker-controlled state, not private actors. this has allowed China to beat the U.S. AI ecosystem recently, forcing Nvidia to join 50 other tech companies in a trenchcoat as a response, in an attempt to catch up to what the Chinese state can do. socialized A also allows the birthplace of China's AI model DeepSeek to have pro-worker anti-AI legal rulings!

Made of

Ethical romantasy

Linguistics is so cool