I’m not very fond of LLM-coding generally speaking myself, but this is an extremely important point to be aware of.
@dalias @torb I can see how bringing up accessibility in an unrelated thread could derail it, but you were already talking about accessibility here. When I give you a firsthand example and the response is "I won’t judge you for using it," followed by judgment of people who bring up why it matters, what I hear is "sure, use it, just keep that relevant experience out of the conversation and shut up, AI bad bad!" That tbh feels a bit ableist to me. We think this tech is bad, and even though it improves the lives of a minority, we're going to tell that minority that that doesn't matter and to stay out of the conversation.
@TheQuinbox @torb It's not a matter of derailment but sowing doubt and trying to convince people working to reject and stop AI to back down.
@dalias @torb I talk all the time about what worries me about AI, especially how much control a handful of companies ran by corrupt billionairs have over it. And yeah, I do want people to reconsider rejecting it outright. Being able to use a device on my own when I used to need a sighted person is a real benefit. Why shouldn’t that count when we’re deciding what to do about the tech?
@TheQuinbox @torb If you're using AI, you're not using the device on your own. You're using it with a malevolent corporation as the go-between in place of a sighted person.
This is the most true when it's an online AI where everything goes through them and they can and will record and harvest everything they process for you - including information about other people who did not consent to that - but it's also true to some degree still when it's a "local model" trained on bad-provenance data, where the biases of the corpus shape what the intermediary tells you.
@dalias @torb Wait, how is Anthropic a go-between when I use my Flipper? Claude helped write the accessibility software. When I use the Flipper, I’m using that software, not talking to Claude. I can pick it up and use it without asking anyone to read the screen for me. That’s what I mean by "on my own." I get this point more for, say, image descriptions, but not hacking a device to talk.
@torb I kinda agree, but there's some parts I disagree with. Now, all vibe coding is slop, I won't back down on that, but I can see their point. I try to make my work as accessible as possible, I would also seriously consider making some accessibility tooling if I knew what was needed(I'm an anarcho-transhumanist so I very much want to use tech to make people equal). And I can see the appeal, but most of it comes from big developers being stuck up pricks not caring about accessibility, from what I can think of there wasn't really any accessibility with emacs until Raman(a blind guy) made emacspeak. A lot of the appeal would be lost with a better culture, I see it as the nuclear option to use ai but that's just me(not disabled, reading for a bit digitally can be hard but that's fixed easily for me with the firefox reader view), like working with the nazis to get stuff to improve your experience while ignoring their victims.
I'm not saying that disabled people should refuse to use anything that even slightly got near genai, use something if you really need to idgaf, but at some point if people start using this as an excuse for themselves(able-bodied etc) to use ai to code then I start to see a problem. It's like when people use disabled people to argue that they, an able-bodied person, deserve to be able to use ai to generate pictures because it stops being about disabled people and starts using them as a wedge, like how the transphobes in my country's government use women's rights as an excuse to do the trans genocide going on over here
I hope all that made sense, fuck developers who don't think about accessibility, they're just assholes and the reason why this is a thing
@torb It's bad faith. The OP in the thread you linked is a known bad actor exploiting disabled folks as props for a pro-slop agenda.
@torb AI does not take care of accessibility needs. It makes an excuse not to solve them by pretending folks who need accommodations can be satisfied by low quality scraps from the slop bots acting as intermediaries between them and the inaccessible things.
If it's a political debate I'd agree: slop isn't the solution.
But if we're talking about real people suffering from the absence of an accessibility device... If they can slop it I hope they slop it. At least until they win they're fight to get the real solution.
So to anyone in distress who thinks slop can be of temporary help, consider it without guilt.
You're use won't change a thing to the success / failure of gen AI : it will fail anyway.
The bet is dozens of trillions invested to replace mankind with AGI powered robots. It's not having a disabled person make an accessibility device to help with something... Especially if that something is so niche the app doesn't exist already.
@torb The tech industry treats this disabled survival as free marketing. They frame themselves as saviors while exploiting the marginalized community they actively helped to disenfranchise. https://sightlessscribbles.com/posts/a-thermodynamic-failure-of-the-soul/