fwiw i'm not interested in applying rigid purity tests for AI use to people. if you've dabbled in AI but saw the fallout around you and decided not to anymore you're a-okay with me. i think plenty of vocally anti-AI people have used it at some point if only to see how the boosters' arguments hold up

it's not about "i have never used AI therefore i am pure and clean and good." it's about people who can see the harms and (continue to) enthusiastically embrace it anyway, that's where the issue is

in reply to rye serling

@rye serling @Eniko Fox

To me it's the insane economy around it. The insane amounts of money thrown at it, CEOs completely losing their minds to waste as many resources as possible on it, often without even having a plan on what to use it for or how to verify whether it's actually good. And of course the push to use it to replace people, replace human-produced content, etc.

Perhaps it's fine to use privately (and preferably from local models), but don't share it blindly with others as if it's in any way comparable to something real. (I have done that in the past and feel stupid about it now.)

in reply to Four Lights!

@nolan I mean, it's not - it's a fundamentally unethical tool. It's built on theft, it's a waste of energy, people were harmed to get thise pointless answers.
The point is not saying it's neutral and fine to experiment.
The point is recognising that it isn't and doing better.
This entry was edited (Wednesday, November 18, 2026, 8:02 AM)
in reply to Eniko Fox

@Eniko Fox I too have used LLMs in the past. The problem I most immediately found was that they give plausible sounding but wrong answers.

The thing I will usually tell someone who relies on AI is this:

Ask it some obscure question about some obscure topic about which you know a great deal, and see how long it takes to give you a plausible sounding but very wrong answer. Then ask yourself if you really trust it with other topics. If you didn't have the knowledge you did, would you have known it was feeding you nonsense?

In my case, I asked it about a small company I had a job at. It hallucinated an entirely fictitious board of directors. That was enough for me to never trust it again.

in reply to Martijn Vos

@Martijn Vos @Eniko Fox Sure, but they typically don't give sources, so "check the sources" typically means doing the research myself, negating the value of the LLM in the first place. 🙃
in reply to Jonathan Lamothe

@Jonathan Lamothe @Eniko Fox
They give sources when you ask for them. Sometimes merely asking for sources makes them correct erroneous claims, but not always.

And if they cannot provide sources, it's probably safe to assume it's nonsense.

This entry was edited (Wednesday, November 18, 2026, 12:43 PM)
in reply to Martijn Vos

@mcv @me there were a few prominent legal cases a while back where even the sources it gave were plausible looking but didn't actually exist

Given that risk it's less effort to not bother with using an LLM and do normal research with conventional sources. Since you have to do that research to confirm the AI output anyway the AI output doesn't actually save any work.

in reply to Jonathan Lamothe

@me @mcv
Worse, they give source-shaped links. If you actually follow the links, you may find something that supports the LLM's claim but is clearly made-up nonsense, or you may find something that directly contradicts the LLM's claims, or you may find that it actually does support the LLM's claim.
in reply to David Chisnall (*Now with 50% more sarcasm!*)

@David Chisnall (*Now with 50% more sarcasm!*) @Martijn Vos @Eniko Fox See, that's the other problem: a lot of stuff online that shows up in a search is just LLM-generated nonsense in a trenchcoat. This makes even doing my own research a pain in the ass.

Sometines I'll inadvertently use an LLM because I did a search for something and the search engine will "helpfully" provide me an LLM response—typically with sysadmin stuff I don't do frequently enough to remember how to do. On rare occasions where that does happen, I'll always consult the manpage or other primary source material for whatever command I'm being asked to execute, but I still feel dirty about it. 🙃

in reply to Jonathan Lamothe

@Jonathan Lamothe @David Chisnall (*Now with 50% more sarcasm!*) @Eniko Fox

True, not every source is reliable, and that reliability is dropping. So you also need to be able to judge the reliability of a source. In a way that was always the case, but AI slop does make it harder.

For programming and sys admin stuff it's still relatively benign, because you get almost immediate feedback about whether it works. There are a lot of topics for which that feedback loop is a lot slower, and potentially deadly.

in reply to Jonathan Lamothe

I asked it to analyze the failure logs at work and create a report. It did (it worked on it for two hours) and it looked very promising. Then I started going through it one by one, asking it to analyze deeper each problem. Most of them turned out to be misleading or only half truths or plain wrong.
So I have to double-check and verify everything it wrote in its 'analysis'. How much time did I save?
This entry was edited (Wednesday, November 18, 2026, 4:38 PM)
in reply to Moah

@Moah @ask i mean, i think there sometimes is..?

to use this example that #CoryDoctorow loves to use: speech transcription. yes, doctorow is using this to make excuses for OpenAI because #Whisper can do that.

but there is also #Vosk (which has been around much longer) with #LibriSpeech models based on audiobooks from LibriVox — not large-scale theft, DDoSing, hyperscale data-centres and slave-like labour for model training.
openslr.org/12/
github.com/alphacep/vosk

#Kdenlive video editor can use both Vosk and Whisper for subtitling. i'd be curious if anyone sees any issues with using Vosk for that..?

This entry was edited (Wednesday, November 18, 2026, 10:01 AM)
in reply to Eniko Fox

if you're wondering what prompted this it was someone writing a thread about how they used AI to explain things to them to learn and then they stopped when they realized how bad AI was. they were afraid prominent anti-AI voices were going to block them for having used AI in the past

and while i can't speak for anyone else personally i'm just not interested in that

This entry was edited (Wednesday, November 18, 2026, 7:50 AM)
in reply to Eniko Fox

Man TBH I have similar fears, simply because my economic situation seemingly forces me to use AI if i want a well paid job. If I was alone, I could prioritize my personal principles here, but I do have kids to feed and there's basically no more programming jobs left around me that won't have me use it rn. But that feels like its already too nuanced for social media, so this might actually alr be self-assisantion. Often feels like I need two internet personas bc of this and it messes me up
in reply to Eniko Fox

one thing I'm always trying to focus on when I discuss the topic with AI users is that we need to bring them to our side. It's not a purity contest, you don't gain internet points or something, it's about steering the world away from an immoral, damaging and unsustainable industry. The same thing you would do with the fossil fuel industry or the defense industry.
in reply to Eniko Fox

Yeah, 'search' is particularly ripe for discussion. From HAL to Ask Jeeves, our collective imagination definitely wanted some form of this. This was how we'd wanted Information Retrieval to work from the start.

I'm still very uneasy about scale (data center madness, GPU/RAM shortages, environment), and privacy (classic, 'Snowden' privacy, and stealing mine, and others, work).

I think laws about training inputs are long overdue, and could implicitly solve both.

in reply to Eniko Fox

I'll put another vote in that column. It's like the gold star lesbian discourse. Nothing wrong with being a lesbian, I like being one very much, but when you get weird about policing what people have *ever* done as part of exploring the world and themselves... okay that's gross. I mean even that was just kind of like... reinventing everything that was weird about hetero virginity obsession. It's all the same stale stuff rehashed over and over. I ain't about it.
in reply to Eniko Fox

I've always been disapointed asking things to LLMs because i've realised I was an expert on the subject and my knowledge were by far beyond the precision of its answers.

Also I hate the sicophant thing, i want to be challenged, contradicted, not flattered. I want my flaws to be pointed out so tha I can improve myself, those machines are not designed to do that.

Instead they make us mediocre and bland.

This entry was edited (Wednesday, November 18, 2026, 8:56 AM)
in reply to Eniko Fox

I ask ChatGPT to describe me every few months, because I know that if I meet someone new in tech who isn't aggressively anti-AI, they'll look me up online and these days they'll probably see an AI summary.

If that fails somebody's purity test... Ok? Everyone is entitled to decide for themselves whether to judge other people... And if so, by what criteria.

Whether I choose to feel bad and/or modify my behaviour based on the judgment of strangers is—of course—my decision alone.

in reply to Andy

@andyinabox soooooooo much this. I hate AI but I literally have no other choice than to use it at work because of the expected output I’m supposed to deliver.

I’d much rather hire a few more humans on my team to do things the right way but that’s out of my control.

So either I use it and keep my job, or I don’t and lose my job. Not a hard choice when it’s that binary.

@Andy
in reply to Eniko Fox

I sit on the "never ever knowingly" side based on science / experience with brain stuff, computer stuff, remote sensing calval stuff. Enough to do an apriori nope and this is why...

Not everyone - in fact hardly anyone - has the same training / toolkit so the purity thing is weird. People learn? Great!

I'm also chronically poor and locked out of professional work - so, I'm really just annoyed that I'm not getting paid for having (and holding to) some strategic vision 🤣 .

This entry was edited (Wednesday, November 18, 2026, 9:21 AM)
in reply to Eniko Fox

Yeah, I'm not allowing any "AI is a sin" talk at all.

I'll use it if I feel like it. I've been playing around with duck.ai just to see what's up. I still end up doing my own search half the time, and of course it seems to promise a bit more than it delivers, but I'm not using it that well. I don't like any of the AI art, either, and the whole thing grosses me out.

AI is being forced into everyone's life, I have no business being uncertain of what it is, or having no ability to make use of it. Nobody does. It's the smartphone again. The people who are about AI are using the hell out of it and teaching themselves everything, while anyone with a purist mindset is getting left behind in complete ignorance. You can't even figure out what it is or your art friend will disown you. Let them.

Artists and creatives have been busy turning everyone online into mere customers for their business. Your art friend is a parasocial relationship that only cares about more Patreon subs, and doesn't consider you an equal person because the Follow buttons say you aren't. Too many artists have now spent years online enjoying what is pretty much a personal autocracy where the people they deal with have no rights and artists only really see the internet as a place to get paid. You have no business in solidarity with them. They want patrons but also want to be the Medici in the relationship.

Nobody has any business letting artists shame them about AI, the artists would NEVER accept you shaming anything they wish to do. People who put "with me or against me" ultimatums on others should quickly taste the whip of "against you" without delay to teach them this isn't acceptable.

It's time for regular people to use whatever technology they see fit to use, as if we have a choice. If the artists hate you, that's okay, you don't need them, they need you. Many of us may need to learn to live without their friendship and carry on.

in reply to James Endres Howell

@jameshowell @jcschue @violetmadder @davidgerard I don't agree with dogpiling or brigading - it's among the reasons I left the corporate Nazi bars - but the way this "AI" booster is being Old Skool flamed by the fedi for denigrating struggling artists, has brought me joy today.

Some folks really do richly deserve it. Very much a "cry me a river" situation. 🤷🏻‍♂️

in reply to Eniko Fox

Purity becomes an issue with people who have authoritarian personality traits to them.

You should be careful with that kind of people. Some research suggests this might be an inborn trait that’s with them forever.

They might never truly change, just adopt new allegiances and loyalties forever.

This entry was edited (Wednesday, November 18, 2026, 9:42 AM)
in reply to Eniko Fox

The media in this post is not displayed to visitors. To view it, please go to the original post.

@Girgias let they who are without sin infer the first stone... or something. It makes sense to try it. It also makes sense not to try it.

(It doesn't make sense to rebuild one's entire life around using it for everything. Though, it might make sense to build around entirely eschewing it... maybe. Like being vegan.)

in reply to Eniko Fox

I am forced to use it at work and it feels dirty every time, but I am afraid of losing my job if I don’t do it :/

If the job market was not so messed up and co-opted by TESCREAL boosters this would be an easy problem to fix. I get excited when I see any news that bubble might be closer to popping, I am keeping my fingers crossed 🤞🏼

in reply to Eniko Fox

yes its normal to check out new things. But when you learn about the damage and bad things you have to stop.
I am sure most people has take a look and just dont say it.
I am against this AI too, but of course I have check it out too. When I was still on DeviantArt and someone has gift me a Premium Member (dont remember how its calls) I had test the AI on DeviantArt aswell and it was totally garbage of course and I didnt even use it for fun and later leave DeviantArt because I dont support this shit. AI dont even give useful references, they was always shit. When you try to use one as reference you realize in the drawing progress how so many things dont makes sense.
As long I have alternatives I will always choose options without AI
in reply to Eniko Fox

I think that it's obvious that a lot of prominent anti-AI voices have never used what they are criticising (in particular around image generation AI) in a way that weakens their arguments.

I had a special interest in DLNNs before the rise of genAI. I'd tinkered with building & training RNNs. When GauGAN first launched I was excited to try it. At one point I tested local copy of Stable Diffusion.

1/?

in reply to Eniko Fox

I went into it wanting to be able to ID vibecoded things better and I'm kind of glad I did lol

I'm okay at HTML and I thought I'd be an okay judge at having it make up a website (It was bad and generic), and then had it redesign my website and it was uninspired at best. It cut some random stuff, like entire paragraphs of text and pictures and it even cut out my footer with nothing to replace it.

I've really only seen a few website layouts that an AI will throw out with some prompting, but I feel I recognize the patterns now so I can better spot it in the wild.

It's like I got the sent and I can sniff em out now :neodog_evil:

This entry was edited (Wednesday, November 18, 2026, 9:41 AM)
in reply to Eniko Fox

Honestly the only time i ever used AI was way back when these image gen tools first started to pop up for public use and only because I was under the impression the people involved were like... not evil and therefore were using only stock images and other free to use assets to train them. As soon as i learned they were just going around the internet absorbing everything (from copyrighted material they have no rights to use to CSAM) that was the end of that.
This entry was edited (Wednesday, November 18, 2026, 12:21 PM)
in reply to Eugene

I should clarify here for full transparency I absolutely never used one of these image generators in my artwork at any point. I only made a few funny images with a super old model of Dall-E (I think) by typing in some surreal nonsense prompts for fun. As far as I was concerned this was a toy someone cooked up and nothing more.
in reply to Eniko Fox

There may even be benefits. With slop agents providing better-than-nothing for the disabled.

Which leads one to have some understanding as to why even sane people might still fall for it.

I’d say these benefits by far don’t compensate the harm, but that’s a next-level analysis that runs counter to instant gratification. And hence is bound to lose the argument. Because a whiff of convenience trumps human survival.

in reply to Eniko Fox

yeah, i mean i also played with like gpt-whatever-it-was-at-the-time and dalle mini to just tell it to generate nonsense (boris johnson eating a lemon whole etc). Chatgpt in particular didn't really appeal to me since now instead of "continue this input" it was "talk to computer" which felt fundamentally wrong, then i saw people treating it as a question answering machine (or using it in serious projects) instead of autocomplete that it obviously still was, so i stopped playing with any of it

after that i checked chatgpt/copilot/etc a few more times over the years just to verify things like:

- wait, does it really get the number of rs wrong? (yes, it does)
- people are saying it's good for programming? ok, let's ask it to write a snippet using an api i wrote myself and know inside out, and see if it succeeds (nope, it made everything up instead)
- ok, so there's no way it will be able to find this thing, but I'm curious what will happen (yep, indeed it's nonsense)
- let's ask it about this thing where nobody knows the answer (and it confidently makes one up)

now I'm obviously vocally against llms/diffusion models because wildly gestures at all the shit happening now but like I'm not gonna pretend that I could have seen it coming back in 2022. At the time i was vocally anti-crypto/nft instead, and I'm very pleased that at least i stopped seeing ape pfps since. Not sure if slop pfps are better or worse, but it's certainly different! (tho admittedly that was also on twitter and I have no clue what's happenign there now)

This entry was edited (Wednesday, November 18, 2026, 1:34 PM)
in reply to Alice

@Alice @Eniko Fox

Yeah, for programming, they're good at stuff that lots of other people have done before. There's tons of examples to be found, so that's how it knows, and that can be very useful at times.

But ask it about something that's new or obscure, and it will sometimes just make shit up. "I don't know" is a phrase it seems unfamiliar with.

in reply to Eniko Fox

What I found interesting is how quickly executive types seem to have gone from paranoid about protecting IP to encouraging their employees to shovel everything in to AI despite copyright on AI generated code being ambiguous at best, and it probably gets ingested for training too...

Can we make our client side JavaScript code impossible to copy? Add this page long copyright notice as a header to every file, that now you have to scroll past every time you open it... We'll have to have the lawyers review that slightly non-standard open source license...

I also don't think they've considered the implications of being entirely dependent on one of a small number of AI companies and what the AI companies can do to the pricing of their services once that dependence is established.

in reply to Eniko Fox

Something that keeps me completely grounded on AI to the point of my hatred is my self hosted LLM server.

Knowing what the pipeline is like from input to output. How to tweak it. How to pre-prompt it. Even what the RAM usage is like between prompts. That all breaks the illusion completely.

That plus things like the Claude Code leak that revealed their "harness" was just begging and that "thinking" is just the bot rattling out everything it possibly knows or can find on the subject before resynthisizing it into an answer.

I am mostly working from an "argue with guys who thinks its conscious" perspective. So knowing where Anthropic is adding the "when they ask if you're alive say you're alive" bit

in reply to Eniko Fox

\\

The problem is AI is such a meaningless word and in many ways just a marketing slogan. LLMs that scrape the Internet and give you answers based on what the majority of stolen text it finds says regardless of accuracy wasting massive amounts of energy and water are just evil. On the other hand I use some photo software that uses pattern matching that calls itself AI to replace unwanted images like power lines with the background, that runs locally on my PC. I also have a 10 year old camera that has an intelligent auto mode that I am sure if it came out today would be called AI.

in reply to Eniko Fox

for me the important questions are, in rough order:
- what social capital + structural power does this person have?
- is this person using those to contribute to the legitimacy and continuation of the bubble?
- how did this person come by their beliefs about it?
- what are their material and ideological incentives?
so to me a professor or a highly respected engineer being a booster, i have a huge problem with that, i've blocked (ex-)mutuals over that.
in reply to Eniko Fox

what is AI in this discussion? Is spellcheck, grammarcheck, TTS, photo editing built into monopolist tech phone OS camera apps, translation, photo tagging, handwriting recognition, swipe keyboards, DSLR onboard face detection, video game enemies, search assist, local special use models, etc. over the threshold of the definition for this discussion? I use FUTO Voice Input Method extensively on a degoogled phone that only allowed me to use dictation years after it was commonplace since I didn't want to send my voice data to Google or Apple. My reasoning was precisely because they were going to use it for training these sorts of things.
1/2
in reply to Eniko Fox

I suffer one where nearly every one of my "10 Blue Links™" that I tap is GenAI SEO slop anyway, so I ran the prompts myself in that case.

In other words, one can't actually "not use" GenAI when we're all within its huge blast radius, so none of us are pure enough to pass our own judgment...

[All non-networked-computer users reading this please write tickled and smug comments amidst yourselves in the paper space provided below. Air gapped users reading, I just pwned your HDD: I'm in it.]

This entry was edited (Wednesday, November 18, 2026, 3:03 PM)
in reply to Eniko Fox

There's accidental usage and unaware usage as well. I'm slowly catching to how AI has taken over the "translate" buttons, which I, ahem, use a lot every day. And then yesterday there was a kanji I could not find anywhere, so I drew it and submitted it to reverse search on, ahem, Google, and the search function was clueless but Gemini woke up (without my asking it to, AFAIK) and pointed me in the direction I needed to go. Impressive but scary, like so much other stuff I'm trying to not need (my phone is an oldstyle clamshell and it's for 911 and AAA strictly).
in reply to Eniko Fox

a year or two back, when it was new, I asked ChatGPT to make me a recipe to see how good it was; it was OK but a bit nonsensical in places so I had to fix it.

I've been meaning to try again to see if it's really improved, just to judge the whole "yeah but the new models are so much better" for myself - because the slop I see shared around is rubbish, but then people are going to share the worst to laugh at, right?

I want to argue with the boosters from a position of knowledge 😀

in reply to Eniko Fox

I did at one point have my Karakeep and Paperless hooked up to OpenAI to do auto-tagging.

It worked well until the day I started to save family birth certificates into Paperless and thought "Does this mean they will become part of a training set? In the future if you ask ChatGPT to generate a birth certificate could it in part be based on my grandma's?" I cancelled my account that day and have never touched any of it since.

in reply to Eniko Fox

I check in every six months or so, and give whatever the new hotness is a fake project, just so I know what to look out for.

The way my girlfriend put it; to convince anyone why they shouldn’t rely on these tools, I need to be able to describe their many shortcomings accurately.

It’s funny because even those brief exposures feel like I’m imbibing some kind of cognitohazard. I absolutely understand how people with a gambling vulnerability might get locked in.

This website uses cookies. If you continue browsing this website, you agree to the usage of cookies.

⇧