AI ruined some of my most precious accessibility tools.
Originally a post on the Fediverse.
CONTENT WARNING: This post contains mentions of ableism, AI, feelings of shame and struggle with accepting my disabilities.
You don't have to expose yourself to things that hurt you. Stay safe. ♥
I used to depend on text to speech and speech to text technology. AI took all of that away from me. It feels like all of these things are AI based now, they all depend on some horrible models that have been unethically trained on stolen data. I struggle to read long series of texts, I have to re-read things multiple times before my brain registers the information, I find reading exhausting and extremely energy intensive...
I don't know if it's the ADHD, the brain fog, having been exposed to so much video based content that my brain has turned to mush or maybe a form of undiagnosed dyslexia. I've been questioning myself about dyslexia for a long time now, but my list of disabilities is getting so ridiculously long that I'm too afraid to bring it up out of fear of being shamed.
But to cope with that, I sometimes asks my girlfriend to read things for me, because I process things a lot better when I hear them. And I used screen readers but they were often painfully unreliable and now, since the AI boom? I find them even more complicated to set up than before, you need to go get your own voice models and so on, and these are all horribly unethical. Getting one that sounds pleasant to listen to is very hard. When Mozilla shutdown #Pocket, I've lost the ability to listen to articles as podcast with a voice model that was really pleasant for me, and I've been missing it ever since, I feel like a core tool of my life was taken away from me.
These new voice models all became locked behind paywall, the ethics of how they were trained dubious at best, blatant stolen voices at worse. They get stuck on words, they fuck up in the strangest ways. Instead of sounding like a robot trying to sound like a human, they sound uncanny and genuinely freak me out. And the whole thing became a pain to set up. I gave up, being overwhelmed by options and the obtuse setups.
Same thing for speech to text. I used to depend on it because between the ADHD and the dystonia, typing is difficult. I often think way too fast and type too slowly to keep up, and I tend to get cramps from typing too long on a keyboard. Typing on glass on my phone sends me into spirals and sometimes full on anxiety attacks regularly because I'm just too hypersensitive and the feeling of typing on glass is absolutely horrible to me.
In some ways, speech to text was better for me before the whole AI boom. What both Apple and Google offered was actually fine, and on my Android phone, I'd use the Google Speech Synthesis thingy and cut off internet access to it, it was fine. Then, "AI" happened. And FUTO unveiled their keyboard, with great speech to text, it was powered by Whisper, but fully offline.
That worked really well but the cracks were already there and I feel like somehow, things have only gotten worse. I was still depending on this until around two months ago where I got fed up with it and stopped using it all together, on a whim. I was sick of it mutilating my sentences, changing words to things that made no sense after having written them correctly, just changing things last minute.
I would send the message without re-reading it no matter how much I tried to think about it, because again, ADHD. Then I would find myself rushing to edit what I wrote because none of it made any fucking sense because of how poorly dictation had botched it. Sometimes, it made so little sense that I couldn't even remember what I was trying to say in the first place and I was forced to improvise. It made it even more mentally intensive than typing, because I would have to painstakingly re-read everything. After multiple passes, I would still find sentences that made no sense whatsoever.
But the most frustrating part and the thing that kept creeping up on me to remind me of the origins of these models, is how it couldn't resist adding a "Thanks for watching!" at the end of all that I was saying. It was a constant problem, a constant thing that I had to delete. Later, it got worse, when it started adding stuff like: "(Subtitles by opensubtitles.org!)" that kind of things.
It frustrates me so much that I've had multiple heated arguements with people trying to defend AI, by explaining to me that "AI might be bad, but it has benefits for accessibility!" Using the very disabled people being thrown under the bus by AI as a shield, when my quality of life has substantially worsen ever since AI turned up everywhere, because I've lost access to all the good accessibility tools that I use to depend on, all of which were replaced with the most unethical and unreliable shit out there.