schneefink | Originally human-created fic please

I've stumbled across "this fic was written by Chat GPT" on AO3 for the first time and I hate that this is apparently a thing now. I also found a post with replies on Tumblr that articulated some of the reasons for that. It feels a bit like, hm, regurgitated writing mush: no thought behind it, no ideas. I think Chat GPT could be useful as a tool for writing sometimes, for prompt generation or brainstorming or something like that, but I don't want to read stories written by an AI. I know that probably at some point soon an AI will be able to read a story that I can't say for sure is written by a real person or not but that's not the point. I want fandom and the fanworks in it to be in conversation with other fans, not translated through a machine copying from other writing.

I also think that out society is woefully underprepared for the current challenges posed by the use of AI but that's entire topics on top of that.

Flat | Top-Level Comments Only

From:

sheron

Yeah that's a no from me. I am not interested in reading stuff written by ChatGPT whatsoever.

From:

schneefink

Me either, at least not as a story. At most maybe as an experiment what it looks like.

From:

yhlee

I honestly don't care as long as it's labeled as such - then people can opt out or opt in as they choose. But I do want it labeled.

From:

schneefink

People can post whatever they want in their private spaces, of course, but I don't think it belongs on AO3. (Which doesn't mean that I want it to be banned, I'd just prefer it not be there.) I don't quite see how it's a fanwork - what's the fannish engagement involved in creating it? There's certainly ways ChatGPT could be used to create a fanwork, I don't doubt there are interesting options, but if it's just a prompt and then ChatGPT wrote something I don't think it's a fanwork or should be on AO3. (And that's not even touching on the question of what the program used to learn.)

From:

yhlee

My take here is the cynical one of: if you let people openly self-label AI-generated without being banned or penalized, then they will tend to self-label with a tag or whatever, and people who don't want it can avoid it (or who are seeking it out, goodness knows, can find it). Or anyway it becomes a possibility.

If you make the policy that anything openly identified as an AI-generated fic or artwork is banned or deleted or whatever, or anything outed as AI-generated is banned/deleted, then people will just lie about it and do it anyway, and then you just won't know (until, I suppose, someone does the detective work to report them or whatever). Also probably a real nuisance to investigate/enforce, and I'm not really under the impression that AO3 is very efficient about its use of staff/volunteers as it stands. I consider allowing self-labeled AI-generated works on AO3 to be sort of a lesser evil.

I mean, I'm someone who posts original work to AO3 sometimes. There's arguably zero fannish engagement involved in that, although for completely different reasons.

Also, in general, I prefer to err on the side of generosity when jurying "what is literature" or "what is art" or "what is a fanwork." If I, a human, create an absurdist text by blindfoldedly throwing darts at a dictionary page, there's arguably zero "engagement" in terms of deliberate semantic/semiotic construction, but I could see some kind of Oulipo argument for it. Likewise, a songwriter might (say) use an "AI" tool to generate chord progressions for their melody - this is a technology that has existed for a while and no one...really blinks at it. There are algorithmic music generation tools where you hit a button, it randomizes synthesizer parameters, randomizes settings on a step sequencer. Heck, one of the earliest examples of aleatory/aleatory-adjacent music is basically wind chimes. is that legitimate creation of art? That said, I realize not everyone shares this stance.

(This is not taking into consideration e.g. questions of legal implications - I'm not qualified to discuss that.)

From:

yhlee

Oh -

melagan's comment below brings up another concern I have, which is that if "discovery of AI use = ban/remove this fic" and there's some rule that you can report this, that's going to be weaponized by trolls against creators who AREN'T using AI, and again, create more work for AO3 in order to enforce this even besides the ongoing arms race in watermarking/detection/etc.

From:

schneefink

Those are reasons why I don't want it to be banned. It would be a huge hassle, impossible to reliably enforce (or even define), and, frankly, not worth it. I still don't like it, and I think it should be discouraged. (Though, not actively/directly because that could turn into harassment very fast, but at the very least not encouraged either.)

From:

yhlee

Hah, crossed threads. Yeah. I mean...some people seem to enjoy doing this for whatever reasons of their own...a lot of people (such as yourself) clearly disapprove. At this point my prediction is that it's just going to keep happening because disapproval isn't going to stop a lot of people in the former category, and possibly become increasingly accepted as (probably) younger people acclimate to it as the state of affairs.

I've been writing dystopian science fiction professionally for the past 20 years, so I'm sort of at the jaded "this is just another Tuesday in the history of humankind." :/

From:

schneefink

I think you're right, there's not really any way to stop it, it's almost certainly going to keep happening. Chances are that it'll even become more common, though as they become more common the backlash and social disincentives could become larger as well, and things with ChatGPT and AI in general develop so rapidly atm that it's hard to make any predictions.

On a larger scale, among the many dangers of AIs like that it's not among the most important issues by far; sooner or later I'm sure there'll be a bigger discussion about AI-written fiction in general, not fanfic in particular.
So realistically telling people to properly label it so it can be avoided is probably the best way to go. I still don't like it and don't think it should be there >.

From:

yhlee

Yeah, it could be really interesting (in the "may you live in interesting times" curse sense) to see how this develops. And, I mean, I'm not unaffected; on the contrary. AI-written fiction (or nonfiction - also big business in traditional publishing) is definitely going to become a thing and it's already being discussed among writers I know in both indie and trad pub; I'm a full-time professional writer making a very good living at this. At some point, for me, probably within the next couple years, it's going to be "adapt or die," because not liking it won't make it go away, so that's what's left.

I do think one of the potential interesting silver lining applications is in accessibility - imagine, as an extreme case, someone who is paralyzed such that they can type with an assistive device, but they are much much much slower than someone who is not in that situation. (I'm thinking of this real-world example). If that (very hypothetical) person wants to save time by typing prompts instead of full fics, and then posting the best fic from their favorite prompt, I'm not sure I would have the heart to say "No, you shouldn't."

From:

schneefink

I can see how it's scary for writers! Probably for technical writers even more so than fiction writers. Cursed-interesting indeed.

Accessibility is a good point, but I think that's more relevant for things like form letters, as a kind of very advanced autocorrect, not for fiction. In a case like that example I still wouldn't consider it the same as a non-AI written story.

From:

yhlee

Huh, I'm not so sure about that! The thing is that this type of "AI" (neural network) has no internal semantic model (knowledge of how the world works). It's all probability-weighting etc. And for technical writing, understanding and clearly, CORRECTLY conveying the content are really important! (Source: my sister is a senior technical writer at Google.)

Whereas for fiction, it's probably genre-dependent...I mean, there are genres like crackfic or niche weird kink smut or wacky comedy where continuity or good characterization or deathless prose simply do not matter to the target audience, so the weaknesses of the AI-generated text simply wouldn't matter. But yeah. It's...gonna be interesting times.

From:

yhlee

(Also - I do realize you said you prefer AI works not to be on AO3, but people's dislike of this is manifestly failing to stop some people from doing so. :] So at that point you're either looking at things like bans/rules, or else at social disincentives, although it's pretty clear to me that the social disincentives have already failed for at least some of the population of people who are inclined to post AI works.)

From:

jadelennox

The most undeservedly arrogant people have completely misunderstood what we wanted AI for. (Also fiction written by a predictive text algorithm is like accidental cereal packet poetry: amusing, but nothing more.

Science Fiction dystopian writers for the last 102 years: "And then we will invent robots and artifical intelligence and the robots will be so angry we make them do all the shitty work so we can live a life of creative luxury that they'll rise up and destroy humanity!"

Silicon Valley: "Ooh, let's invent algorithms to read the fiction. And write the fiction based on the fiction it read. And look at pictures. And then make new pictures. Somebody send the proles to the Uber-driver mines so we can spend more money on our galaxy brained vision."

From:

schneefink

fiction written by a predictive text algorithm is like accidental cereal packet poetry: amusing, but nothing more.
Hah, yeah.

It keeps being frustrating what things that could make our lives so much easier are actually used for. (Though off the top of my head I struggle to think about what things I would prefer that AIs like ChatGPT or its image equivalent would be used for, but I'm sure there's something.)

From:

michelel72

One of the fic comms recently posted a policy about AI-"written" "works", because the question had already been raised. I sincerely don't get the point of posting something spat out by an algorithm.

From:

schneefink

And what policy did they decide on?

I guess maybe if ChatGPT writes someone a story for a prompt that they really like and think other people might enjoy it too? Still doesn't feel like "fanfic" though. Some kind of imitation.

From:

michelel72

It looks to me like "contribute some work of your own" and "label it". Post is here.

From:

yuuago

Using it for a prompt is one thing, but posting the whole thing that it spits out... Ew. No thanks!

From:

schneefink

I don't want it to be called a fanwork, especially. If it spits out something amusing that might be something worth sharing but it's different from fanfic.

From:

muccamukk

I guess it's nice they label it so you can not read it?

Honestly, my main concern with these bot driven text generators is the stories about the underpaid labour used to screen the content.

From:

schneefink

Definitely better for it to be labeled!

Overall, yeah, that's definitely a major issue, and the danger of the "advice" it can give and that people way overestimate the accuracy and reliability of its responses, and then there's also copyright issues.

From:

hokuton_punch

Yep, that's about where I am on it. :/

From:

schneefink

Overall I'm more concerned about other things in regards with ChatGPT - I think suggestions that it could be used in legal and/or related fields are not only misguided but actually dangerous - but I find this annoying too.

From:

merit

It seems to be rapidly missing the point of fandom.

From:

schneefink

I guess if I try to interpret it charitably as wanting to share stories you enjoy? But, even disregarding the question of what exactly the AI is copying, for me fanworks are things created by fans, and this does not qualify.

From:

marginaliana

I hadn't heard of this happening until this post but now I'm so angry about it. What do people think they're going to get from doing this? What do they even want? This is not fandom.

From:

schneefink

I'd heard about it on Tumblr but I'd hoped that it was a very niche case, it was an unpleasant surprise to find it myself. Maybe/hopefully I just got very unlucky and it is still very rare.
Trying to come up with a charitable interpretation, maybe they enjoyed the text the AI wrote and think others might enjoy it too, but yeah, that's not a fanwork for me.

From:

melagan

I received this comment on a recent fic.

Comment #5K580 NovelAi use ́in this work has been flagged. Readers, thére are AI detectoŕ tooĺs ́like gowinston that will detect any AI use.

I definitely did not and will never use an AI for my writing so I have no idea what prompted it.

From:

schneefink

Ugh, that's annoying. I read that there's spam comments like that going around randomly, I'd just ignore and delete.

From:

china_shop

In case you aren't aware, apparently there's a spambot leaving comments like this. (Ugh!)

From:

china_shop

I'm personally very DNW about people posting AI-generated stuff as creative works (whyyyy?), but I did briefly go down a rabbit hole of thinking I could use ChatGPT in private for language learning. There's a dearth of graded readers in Korean, as far as I can tell, so I asked it to write me a few short stories in intermediate Korean. The results were... not good. Trite and empty and very summary-ish. No characterisation. I had thought that if the point of the reading was the language learning, maybe I wouldn't mind, but I was wrong. /another datapoint

The other thing I found was that, because there's a potentially infinite supply of this stuff and 99.99999% of it is crap, there was no incentive to focus on the text in front of me because there's definitely a better possible text it could create -- so it turns into a kind of slot machine situation.

(Here via my reading page; hi! :-)

From: (Anonymous)

Just had a conversation about this yesterday, and how it doesn't count as art if it's not genuinely expressing anything. Even a terrible fic you hate was at least created by a person who had something to say and share.

That said, I expect this problem will solve itself once the creators of ChatGPT decide the tool has been polished enough and inevitably make people pay to use it.