#gen ai discourse

[mdconfession-blog]

it's frustrating seeing unlabeled ai-gen music get big in the murder drones fandom. its obviously ai and you can't point it out because "ai" and "suno ai" are blocked in the YouTube comment sections. the associated fic is also likely ai :(

its scary. people work so hard to hone their skills out of genuine passion for their craft only to be overshadowed by a cloutchaser with ai. it's not just murder drones, but this is the fandom i've seen those ai folks be most successful in

⠀

[radicalcrashout]

maybe i'm insane but this is just genuinely so sad. like truly and genuinely sad i feel bad for her it isn't Fun to be addicted to social media, it isn't Fun to call yourself "maybeimnotsmart" online (idk if that's made up for the article tbc and idrc its definitely real for Someone out there), it isn't Fun to feel like you don't know how to do anything at all without the help of a machine. it sucks. and i do genuinely believe we can recover from this but we do need to start addressing being Actually Addicted to your phone because of how life is and how we can work our way back out of it

(https://archive.is/7yQqX is the article in question im new to tumblr i dont know how to link things properly i hope thats ok)

[elf-trash]

reposting this bc the OP blocked me (and is blocking anyone else who disagrees which means blocked people can't reblog) and i want to say this loud and with my whole chest!!!!!

another Dragon Age fic was recently outed as being AI, and this is what the writer had to say for themselves about it:

so actually, Grammarly uses generative AI and is just as bad as ChatGPT. it also objectively makes your writing worse, it sucks the voice out of your prose and turns it into corporate sounding homogenized paste. it's also unethical for all the same reasons any generative AI is unethical. get a writing group and have a real human beta read for you if you don't trust yourself to check your own grammar etc. but honestly something unpolished and written entirely by your human brain and human imagination will ALWAYS be better than AI slop.

also, the part about published authors doing this is patently untrue. i know this is a huge problem in the self-publishing space, but most publishers now are including clauses in their contracts that expressly forbid the use of AI in ANY part of the creative process. this includes using ChatGPT to generate or clean up outlines or Grammarly to spellcheck and revise. so if you're trying to publish, don't fucking do this or you could literally be asked to return an advance if you get caught.

i've posted about this in the past, but AI detectors are actually shocking accurate these days. i've tested them extensively recently and they can consistently and correctly flag individual sentences written by ChatGPT in an otherwise original passage. and they almost never flag false positives. so the argument that AI detectors can't be trusted is just flat out wrong. are they correct 100% of the time? no. but can they indicate with a high degree of accuracy if AI was used in some capacity? absolutely, especially if there is additional evidence.

and for all the people hand wringing about AI detectors flagging false positives, let me just say this: if something is not AI written it is very easy to prove. you can't write anything of any considerable length without leaving a massive paper trail of notes and drafts. almost all writing software tracks changes and makes it very easy to prove you wrote something yourself. being falsely being accused of AI isn't actually a real problem and is only being made to seem as such by people who are trying to get away with and justify using AI or who are worried about getting caught.

i think a lot of people are just lured by a seemingly easy shortcut, and to their untrained eye, what the AI is spitting out feels "better" to them than their own writing. but i promise you it's not. trust your own brain and put in the work to improve at your craft rather than outsourcing the gift of your imagination to a robot that steals from other people's work.

i will continue to die on this hill!!!!!

[garlicbreadslice]

“I use gen ai because I want to make art but I’m not good enough at it to make the art I want :(” well friend guess what. You’re still not making art. And you’re still not good at it. So like. What did we learn here.

[kimyoonmiauthor]

AI Bros, yo...

So someone tried to say because I was against Gen AI, that I was being "Elitist" lol, "Ableist" lol and "anti-poor people."

Dudes, are you even disabled, poor, or have grand privileges like being a white male (they had a picture of a white male on their profile)

You're against Poor people

Me, who was given up for adoption and still has trauma responses from knocking on doors because I was so dirt poor, they had to give me up for adoption. And me, who used to walk to the library because I didn't have internet on my free days to get air conditioning, to submit stories, is being effing prejudiced against poor people. Oh, how rich it is to say shit like that. And you know what? I still drew and wrote and submitted stories.

But, but you know that I am being really prejudiced against poor people. Do you want to talk about that?

I used computer paper and a pencil. And have you seen the artwork of Slay the Princess which is pretty much almost all mechanical pencil?

You're being insulting to poor people. You're saying poor people can't do art or be creative. When backed into a corner, 100% I was great at art. I made art with brie, I made art with a rock I found on the sidewalk.

Lascaux's caves is basically chewed and spat out dirt, dudes. And you're saying you need AI if you're poor. WHA~~ Some great artists were poor their entire lives.

Rembrandt???

I've made art out of an orange peel. Art is a result of resourcefulness and creativity.

You hate NDs

C-PTSD, dyscalculia, SPD. Yep, I hate myself. And damn it, I hate NDs, and fuck, I can't draw anything. OMG, I need AI /s

This is the only AI I used. This is called Adobe Illustrator. [A]dobe [I]llustrator, though I used Affinity as well.

I did test AI mostly to mock it, but I clearly marked it.

I created 100% of the brushes and all of the vectors. "OMG, I can't do art. I have a disability that makes me can't do art." /s

I heard the Autistic excuse crap too (that person wasn't autistic, BTW, I checked). I've done art exchanges with people in that ND camp as well, in fact, sometimes I feel like they are better than me.

You hate Disabled people (some of the bros called everyone else The disableds... ?.? And then I know you're not disabled)

I have CPTSD, which is more like a cocktail. I also get chronic pain from issues you don't need to know.

And I still was able to draw this. OMG, you're terrible and hate disabled people when you insist on not using AI.

BTW, There is no disability around creativity.

Disabled people don't have a disability around creativity. I saw someone sew clothes with their mouth. Are you going to tell me that sewing isn't creative? Someone was paralized from the neck down, on vents and was typing a novel from a computer tracking their eye movements. And this isn't the inspiration porn moment, but just saying. If they want to, they can.

You're being Elitist

Huh? lol rotfl

I have a degree in Anthropology concentrated in systems, I'm likely more of a red shirt. You adopted? You queer? You have disabilities? You're ND? You're female? You got called not real because you're all of these things and called a catfish?

You, gen AI bro, are more likely to be more elitist than I am.

And then he slinked off...

Because he had no defenses, and couldn't figure out how to think independently without AI. Whereas, I swear, the majority of my knowledge was gained by going to libraries (Some of that work is on this blog, BTW with clear pictures of the books I collected. Look for Robert Goheen. I took that picture in the university. I had to go to a university library for that. And no, I did not attend said university, but it is allowable by university rules as long as I don't try to check the book out.), which, BTW, he also called elitist. HAHAHA. I pay no money for books and actually put in the effort and that's elitist?

The majority of learning how to draw was done on my own by leveling up. Yes, I did take Graphic design classes. Yes, but most of that knowledge is now on the internet. The advantage of classes is being around other like-minded people and having that collective energy push you to be better and find out what you can and can't do, to push your limits. And I'm sorry, no AI can do that.

The best AI can do is fuck up hanbok and not understand what time period the hanbok came from. And I'm not going to teach it that.

And if it's elitist to say I clawed tooth and nail to be able to draw and spent time and effort, then fuck it, anyone that wakes up in the morning and puts in daily effort to get better at anything is elitist for not being lazy and learning how to think for themselves.

[nebby-the-protogen]

the whole "ai steals art" pillar of the anti-ai moral panic is just a game of broken telephone mentally justified after the fact

[moran-with-a-g]

Stop giving ChatGPT so much credit.

It's not amazing, it's not horrible, it's literally just a tool.

It has some uses where it's good, it has some uses where it's bad.

The whole thing about it hurting your ability to think depends on how you use it. Does TikTok not shorten your attention spam? Do computers not make you go out to the sun less? Did they not claim TV ruins your eyesight? Did they not claim using a calculator will make you dumb?

I had to write a ton of regex this week for a monitoring dashboard I was building and you can be sure as fuck I used ChatGPT for that, and I actually started understanding how regex works and started needing it less (even though regex is still evil). Because I also took the time to understand why it worked, and didn't just blindly copy-pasted. I did not use ChatGPT to write any other part of the project, because I didn't need a tool to generate the descriptions for the graphs, or the layout, or the metrics. But if I didn't use it for the regex I would have done maybe a third of the work I managed to do this week, and for what?

Should you use it to write whole stories, essays, etc? No, I don't think you should. That's not what the tool is for.

But if you're writing a fanfic and you want some cool CSS to make the text shake for some part of it? Yeah you could generate it using chatGPT.

And even if someone uses it to generate a story or an essay, they can do it in a way that still challenges their thinking. They can use it to generate points to work from as a base. They can use it for a first draft and then rewrite it. They can ask for a metaphor and work from there.

And if they don't? That's on them. A nurse is not going to pass medical school only using chatGPT, there are practical exams they'll need to pass. And how is that different from normal cheating? You think people didn't cheat before ChatGPT? You think there aren't tools that help you find WHO is cheating with ChatGPT? It's actually easier to spot than normal cheating.

Y'all are making such a fuss about it. Don't like? Don't use it. But don't shame people who do. If you don't want to see AI generated stories on AO3, stop shaming people who post them so they start tagging them properly so you could filter them out, or know not to click. Have an issue with people who cheat using chatGPT, because it doesn't engage the brain? Show them the correct ways to use the tool to actually engage their brain. Encourage them to try and understand the logic behind the results they get.

And maybe focus about targeting the companies behind the GenAI instead of the users, to make the tool less harmful for the environment and to stop using stolen data to train it? Or, you know, develop an alternative that only uses data that was given with informed consent and that the creators were compensated for?

[cannimumsable]

the "disabled people need it!" argument for generative AI is insane because it completely misses the point. like yeah, i get it, it could be great if disabled people who genuinely cannot bring themselves to type or hold a pencil could have a way to create anyway. I really do understand why that's an attractive idea, I am aware that some disabled people just can't draw or paint or write and it's not their fault and there's nothing they can do about it. That doesn't make gen AI use ethical. The problem isn't that your art becomes soulless and Bad because you didn't put your blood sweat and tears into it (I don't like gen AI art for this reason too but I am aware it's subjective and don't use it as an argument), the problem is that generative AI is killing the environment, stealing from artists who never gave their permission to have models trained on them, and the democratization of gen AI is just asking for misinfo and propaganda to be more common (it's already bad enough).

[fictionyoubelieve]

Anyway, I’m using generative AI pretty heavily at work these days, and it’s been great! It’s like a smart rubber duck, specialized documentation search, and advanced autofill, all in one. It has its flaws, but you learn to work around them, like with any tool.

The second part is particularly useful because web search sucks so bad these days and doesn’t seem to be getting better. I do worry that access to good documentation sources (including first-party docs, but also things like the Stack Overflow archives) might become more gatekept as it becomes more valuable as training for these tools, which would be to the detriment of all. It already happened to reddit. Probably a lot of that and the web search degradation is due to ZIRP ending and general profit-motive enshittification, but I still suspect that gen AI exacerbated it.

I cannot find fault with the tool itself, though. It’s been helping me a lot, and I enjoy using it. I’ve really been blown away with how much it can do already, and it seems like there could be a lot of improvement still coming.

Where it’s at right now, I definitely don’t think it can be a 1:1 replacement for even junior programmers, but maybe the productivity boost could reduce overall demand a bit. I do think it’s really effective as a training and teaching tool, so I’m guessing it will somewhat reduce barriers to entry to the field, which would be a good thing.

[whyamiherewhosummonedme]

"based on your likes!!!" and its fucking AI discourse

[library-seraph]

Nothing makes me feel quite as out of touch with reality as coming on here and seeing people play the "being anti-ai art is just like believing in the concept of degenerate art" card and then going back on the former wretched bird site and seeing a ton of the roman statue trad dipshit accounts heralding their gen AI oil paintings as being what will save western culture from lgbtq artists and modern art

[elf-trash]

my sister just made this banger post and i had to share it here too

[nebby-the-protogen]

i'm not denying it's happening but this is the venn diagram

(ps, i don't mean "replaced" in the same way piracy retroactively "steals" profits. i mean people who are actually losing jobs. not some guy generating an image of his furry oc to have a quick jork to "instead" of paying someone to do it)

[22degreehalo]

It's weird to me that the same people who laugh the hardest at anyone who claims AI is 'infallible' also seem to, like, take that as the only possible goal AI could strive to?

Like, the mistakes generative AI make are incredibly similar to those that humans make!: finding just the first search result and not looking any deeper, too easily getting bullied into giving in even when they were right, responding more favorably to people who speak to them politely rather than rudely, etc.

Lately, I've been using Claude and ChatGPT a bit more primarily for cases when I'd look up a subreddit and ask a human there the same sort of question, and I've found it about as useful and accurate as that, just without the necessity of an actual human going to the effort. Sometimes it's wrong, but sometimes humans are wrong, too?

And it just makes me think so much about how eyewitness testimony is by far one of the least reliable forms of admissible evidence in trials. That if a computer was wrong as often as eye witnesses were, it'd be grounds for a mistrial even to submit it.

And...I guess: I just don't believe there is any definitional difference between a human brain making an analysis and a computer doing it. Yes, human brains are far more complex, but that's a matter of scale. I do not believe in souls. We are, fundamentally, made out of the same physics and mathematics as AIs are. (Also, I'm a free will sceptic, so... there's that.)

I get that there are good reasons why humans have this sort of pro-human bias, but... it very much is a bias? And it certainly seems that it can and does hold us back or lead us to all kinds of bad outcomes. Like, bad eyewitness testimony could lead to someone being jailed or even executed despite being innocent. That is a very bad outcome.

I guess it's also a loss aversion thing. If AI leads to something bad, that's something that could 'easily' have been avoided, and needs to be publicised heavily. Whereas if it leads to a good thing? Okay!!! We weigh these things heavily skewed towards the negative: if a change has even a few bad aspects to it, we tend to prefer no change at all, even if overall things are better. We view human error as something unavoidable, but AI error as an 'easy fix', and AI good as a simple 'bonus', if that.

And I don't think that's fair. There are plenty of ways AI can be helpful to anyone - not just 'capitalists'. (Honestly, given how open-source AI is, it's extremely weird to me that so many people have even made that connection. Like, if anything AI has evened the playing field between corporations and individuals?) Any mistakes should be judged in light of this, and in comparison to how often a human in the same sort of position would do the same.

[nacricissa]

If gen AI was just a tool that got used like one. I would be fine with it. If it was a way to take large datasets and produce similar content (which it is) and it was treated like that (which it isn’t) we’d be chill, LLMs and I. However, it is existing in a situation where it’s built of theft and lies and is being treated like a whole person that knows facts and that’s not what it’s for.

Helping my dad rephrase his search query to get at what he means but can’t articulate, directing him towards the Actual Human People who’ve worked on this problem? Honestly would be great if it hadn’t been built in a way that exploits both labour and the environment.

Translation! LLMs are a nice tool to help people communicate in a language that isn’t theirs which would be way nicer to think about if the language wasn’t also stolen from its original creators.

Similar tools can be applied to data analysis in big data, like bioinformatics or astronomy in a way that dodges a bunch of the gen AI pitfalls but we call all of these things AI which makes it impossible to talk about any of it without people’s love of Wall-E or Baymax getting incorrectly triggered.

I’m starting to sound like a nutcase at work because upper management keeps trying to implement AI programs and AI assistants and Chat GPT and my middle-of-the-road, don’t-infodump, don’t-engage response has been “I don’t like AI”, “I prefer to remain in control of my own tasks”, “I’d rather make my own mistakes”, and “I don’t trust any machine smarter than a toaster”

[tanadrin]

i think one of the most interesting things about generative ai is not just that it was a pretty unexpected thing--seems like very few people were sitting around ten years ago imagining we would have this technology in 2025--but that i think it is also pretty difficult for people who aren't well versed in the technical background to trace how we got here from there, you know? like when the internet became a big thing, i think if you were familiar with the concept of the telephone or even just one computer networked to another somewhere else you could grok the fundamental concept: it's just a bunch of electronic machines connected to a bunch of other electronic machines; it's an extremely cool piece of engineering, but packet-switching is not (at least at the nontechnical level) that conceptually different from a telephone exchange.

and you could extend this backward pretty far. electronic computers from mechanical ones; the telephone from the telegraph. likewise future developments that emerged from the internet: smart phones are not to conceptually different from computers and radios, they just ("just") are very sophisticated devices that use new versions of those older technologies. and a lot of technology is like that. if you understand a cannon you can understand the basic principle of the space shuttle.

gen ai seems... not like that? that kind of, i guess, statistical approach to problems in computer science wasn't invented in the 2010s, i gather it's a lot older, but it was mostly a niche research topic, i think? and there were some nifty demos of still pretty crude versions of stuff like deep dream, but it's not like we'd had twenty years of this kind of stuff being part of the wider milieu of technology in everyday use before gen ai started getting good. it's weird! it wasn't an accident, people had been working on this stuff for a while. but in some ways it feels like the discovery of antibiotics, one of those medical breakthroughs that happens just as kind of an a priori discovery of something useful out in the world.

and because computers are already omnipresent in our lives, unlike a medical breakthrough, it's suddenly everywhere. and yeah often it's used or promoted in ways that are pretty obnoxious, but even still, no wonder it provokes feelings of dislocation and anxiety. technologies which emerged much more gradually into society have provoked just as much unease. and the idea that it might keep getting more useful, as much more useful as computers have gotten over the last, say, 25 years--that's just hard to fathom from any angle. i think it's as hard to estimate what kind of social impact that would have as it would have been to anticipate all the social impacts of the internet back in the 1980s.

and it kind of seems a pity to me that the three camps in the discourse right now generally seem to be "ai is useless and stupid and a fad and a scam", "ai will destroy the human race", and "ai will usher in a post-scarcity utopia," because the possibility that ai is neither a complete mirage nor the end of human civilization as we currently understand it is much more interesting. and much harder to speculate about.