brib ​

#AskFedi  I came across a study a while ago which had considerable evidence that LLM autocomplete can shift the views of users without users realising their views have shifted. I want to talk about it in my upcoming essay. I can't find this study; DuckDuckGo and Google Scholar aren't helping. Can anyone help me dig it up?

saying ai is inevitable is like saying the monarchy is inevitable

@aisling@critters.gay Given it's really difficult to tell LLM images from human images nowadays, the source is one of the better ways to determine if the artwork is human-made. A lot of genAI spam will come from sites that value quantity over quality while a lot of human works will have an actual artist with a distinctive style and a history showing a progression of work.

@zzt@mas.to Oh so now Linus is the guy going "it's a tool and its usefulness is undeniable". So, I should point out, is asbestos

@baldur@toot.cafe Slightly off topic, but this reminds me of something... Something that isn't talked about enough in the AI and disability discussion is how many creative careers (and software development too) were taken up by neurodivergent people who could take advantage of the flexible hours and work from home opportunities. These are often quite important accessibility needs that aren't readily imparted by employers. Nowadays if someone wants a piece of work flexibly and via a computer, they can prompt a chatbot for a "good enough" result. The result of this is while some disabled people are talking about AI's accessibility potential (e.g. for describing images), other disabled people (including myself!) are finding themselves in an extremely precarious position thanks to AI displacement

@davidgerard@circumstances.run If I had a nickel every time an AI critic got a sponsorship offer from an AI company I wouldn't have enough to buy anything especially useful But I would have enough to question why all these AI critics are getting sponsorship offers

@bel@terberlo.dog I had to check the alt text to understand what this was about Take my groan-filled reboost :neocat_facepalm:​

For #Codeberg members: In the wake of finding my software in The Stack I'm going to raise the issue of requiring FLOSS licenses when FLOSS licenses grant access to any use-case, when I (and I suspect other devs) want to be able to use licenses that forbid AI training. The discussion is at https://codeberg.org/Codeberg-e.V./Discussion/issues/195 I don't think it's available to non-members. I want to turn it into a formal proposal and don't quite know how. I would appreciate help in this area. #noai

@grimalkina@mastodon.social I don't have a huge amount of experience in this area and most of my chronic illness diagnosis journey was done via Dr Google / social media (which also got highly derided at the time). I'm pretty anti-LLM but I'm also aware there can be complexities and I don't want to go about yelling at people when they're e.g. in the position of the 50 year old cancer patient and just want answers. In your experience, when it comes to harm (e.g. going down a wrong information rabbit hole) vs benefit (e.g. the benefits to motivation), how do you think AI use compares and fits in with these older methods of using the Internet to figure out chronic illness? Are they similar, or do some methods have better benefit-against-harm profiles? Has LLM use affected the ability to access non-LLM methods? (I'm thinking how it used to be easy to access human-written information but now it's much harder).

@ainmosni@social.ainmosni.eu I will have you know even the cartoon supervillains in our system don't go that far

@moriel@chaosfem.tw @ldottxt@kind.social everything is more expensive when you're disabled for example, additional cost of healthcare, living expenses, accommodations, bad financial decisions if you're blessed with adhd, etc