I’ve found that LLMs spit things out that read like bad high school essays. I’m not sure they’re succeeding at sounding allistic at all. Just weirdly repetative in the way a structured high school essay is.
Yeah it’s tricky to make it sound normal. Esp since normal very much varies based on setting and culture
I’ve spent 2 months transcribing an entire poorly written text book into a Google doc. I’m now taking that transcription and having chat gpt rewrite it all for readability. All so I can maybe pass certification exam.
The problem is less with us and more with academia having developed an highly oppressive way of writing things. But from my perspective it’s just sloppy unreadable garbage.
AI has been great I can just give It the promt “make this concise and readable using only common language” and it will take entire chapters down to simple point form lists for me.
I also use goblin tools for writing.
I do not know goblin tools… but soon I will
Just in case there are malicious clones of it out there, the official site is goblin.tools 👍
I went, I REALLY liked it.
I haven’t had a reason to use it properly yet, just testing, but it looks great. It’s not perfect though. I asked it for a way to check what programs I have installed on Windows because I want to switch to Linux, and its answer was that I should make a list of the installed programs >.<
Tips for switching to Linux cause why not:
-
Don’t… until you’ve switched to cross-platform software that you know fullfils your needs.
-
Use a live session for a few days, weeks, months, before you actually install it.
-
Don’t expect it to act like Windows, it isn’t
-
As someone who doesn’t have adhd/autism, I see this as one of those legitimate uses of AI. Because lots of people struggle to make mails/texts to be readable.
Other is summaries the context of large texts/data sheets.
Claude has the ability to accept uploads of PDFs and then answer questions about them. I just recently uploaded a PDF of a complex state tax law that had deleted portions and addendums and all kinds of stuff over 36 pages of legalese, and after a few test questions to see if Claude was able to do what I wanted, I started asking questions and getting answers (with references) for important things I needed to know about what my rights were, and what I could and couldn’t do.
I never could have figured that out just reading through those pages.
@Melatonin @plactagonic just be sure to double check all references for hallucinations
That’s a great reminder for everyone!
I never trust AI. Or rather, trust but verify. In the case of a PDF I ask for section, page number, or quote. Always get the reference.
@Melatonin Yeah, I usually run into people who either assume everything is a hallucination or don’t understand that hallucinations happen and are unavoidable. Even less people even understand why or how they happen (ie. if you ask a question about anything not in the provided info, it’ll most likely hallucinate as if the answer was there)
I’m actually still coping with the fact that I can use AI for work. I hate it. It feels like cheating and I learn little from using it vs. figuring the thing out myself, but this is a smart use.
@finkrat @Melatonin if you read it’s output and learn nothing then you’ve only saved yourself time, if you don’t understand and learn nothing, then you shouldn’t use it because you can’t vouch for it
I don’t have autism, but I still use ai to write out boring corpo stuff sometimes. Like out of office replies and sometimes to add some structure to an argument I’ve typed out hurriedly etc
This is literally how I’m getting my directors to stop pestering me about how complex my shit is. Dumbing it down and translating my messages for them. Works wonders.
I sometimes use AI as a proofreader. Asking if the text is well structured and how I can improve it. I prefer to rework it by myself, but it’s nice to be able to get a feedback on a report you are writing before sending it.
But my main use is to ask « common sense » things or fill my lack of basic knowledge. For example, I was struggling for buying some honey at a store because they were 3 kinds of honey and I had no way to know which one to buy (it was a bigger a store than the one I usually go to). I had a short conversation with an AI to determine which one was the best for me. It calmed me down and helped me to make the right choice (this is the kind of situation that makes me very anxious).
It’s also very good to learn or understand foreign language expressions. English is not my native language, so it’s nice to be able to ask an AI about a joke characters are telling in a RPG when the game has not been translated.
I think the next step for me will be to give an AI the ebook I am reading, and ask it questions about things I forgot or did not understand correctly while reading it (I don’t want a summary of the whole book because I don’t want to be spoiled).
I think it’s a very useful tool, and I believe it could make a big difference for autistic people as well in some cases.
I’ve been great at writing since I was a kid, so I hadn’t even considered it since writing is the one place I can express myself properly.
Not necessarily ai, but I use recommendations of message app to make a 2-3 word answer instead of a single word one
I do lists of stuff I want to say, make ai do it, then change it to sound more like me
I use it to save energy for meaningless texts.
I’ve been loving pi.ai for this, it seem to have far more emotional intelligence than any others I’ve tried
I tried it and I really liked it!