![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
* AO3 I love you, but you are really fucking up hardcore when it comes to AI. Also, these are unforced errors. Having never touched the topic would be better than the shit they are saying.
"If fans are using AI to generate fanworks, then our current position is that this is also a type of work that is within our mandate to preserve."
No, it's not! Despite all the concerns I have yet to hear about any specific fandom having an AI spam problem, now they are going to. This is going to be seen as an invitation for people to try to take their shot at getting relevance and a voice in their fandom. AI generated works should be in the same category as plagiarized works. Also, it's AO3's job to preserve the work of fans, not fucking plagiarism bots being handheld by exploited labor in the global south.
* I went for a long walk last night. No pics, but it was good to do a longer walk again. I kinda need to build my ability to walk that far back up. I've been having heel problems I am trying now to aggravate, but paradoxically that means I sit more which also causes problems. I am really good at walking nearly every day, just 20-30 minutes, so that's good but I need to do more.
"If fans are using AI to generate fanworks, then our current position is that this is also a type of work that is within our mandate to preserve."
No, it's not! Despite all the concerns I have yet to hear about any specific fandom having an AI spam problem, now they are going to. This is going to be seen as an invitation for people to try to take their shot at getting relevance and a voice in their fandom. AI generated works should be in the same category as plagiarized works. Also, it's AO3's job to preserve the work of fans, not fucking plagiarism bots being handheld by exploited labor in the global south.
* I went for a long walk last night. No pics, but it was good to do a longer walk again. I kinda need to build my ability to walk that far back up. I've been having heel problems I am trying now to aggravate, but paradoxically that means I sit more which also causes problems. I am really good at walking nearly every day, just 20-30 minutes, so that's good but I need to do more.
no subject
Date: 2023-05-13 11:16 pm (UTC)From:no subject
Date: 2023-05-13 11:55 pm (UTC)From:no subject
Date: 2023-05-13 11:33 pm (UTC)From:Though, there is the problem that if you ban AI generated works, people will still post them but won't tag them, making them more difficult to avoid... It's a tough problem. I know that I wouldn't be sad if I never had to see an AI generated work on AO3, but I also know that banning them outright probably wouldn't do much except make people not tag them. Right now, people seem to be willing to note them as AI generated, which for me is a big blinking neon sign telling me to take myself somewhere else.
It's a no-good situation all around.
no subject
Date: 2023-05-13 11:59 pm (UTC)From:Realistically, there are a lot of rules the AO3 can't enforce. There are multiple fics that have plagiarized me, just because nothing can be done doesn't mean we shouldn't have policies and stances against it.
Despite worries of Ao3 getting spammed, it wasn't really happening. Communities were somewhat self regulating on this. They really should have left the issue alone, especially since it's such an evolving thing.
no subject
Date: 2023-05-14 03:22 am (UTC)From:no subject
Date: 2023-05-14 03:49 am (UTC)From:The policy has always been that you cannot post works that do not belong to you/that you yourself did not make. You can't steal another fan's fic, you can't find-and-replace names in someone else's work and call it yours, you can't repost that one fic you saved from a fansite twenty years ago even if your only desire is for it to be preserved.
AI work is trained on stolen data, and even if you want to argue that it "remixes" that (stolen) data, that's secondary to the fact that the person posting the AI-generated work did not actually create that work themselves. Even if the AI itself isn't going to fight to claim attribution, the work was not produced by a fan, and is being posted by someone who did not create the work in question.
And no fucking kidding - "well, I mean, if the spamming gets egregious maybe we'll step in" is just an engraved invitation to try and find out where the line is!
I mean, I can hope that AI spam won't ever take off due to plain community hostility/lack of monetization, so the rewards for spamming a fandom or a ship or a tag won't be worth the effort.
But still, it shouldn't even be a RISK that a real fanwork can be pushed down multiple pages of results simply because someone else can choose to post a ton of AI generated crap. And let's be real - that's a PRIME choice for harassment. Generate a bunch of short "fics" to clog the tag of that ship you hate to frustrate the people who do like it. Post a ton of AI garbage the instant that author you hate posts a new chapter, so they don't show up on the recent works page.
I won't ever choose to read AI "fic", but I don't want my search results to be clogged by it. I don't want to have to click through on a fic that sounds interesting by summary/tags just to discover that it's actually AI garbage.
And despite claims to the contrary it IS different than the forever-problem of clicking in and finding a fic that has bad spelling/grammar/plotting/character etc. Bad fic happens, and it's a bummer when you have to nope out of something because you don't like the execution, but this isn't just that.
Ugh, I hate physical issues that seem to be made worse both by doing things and by not doing things. Unfair! I'm glad you got in a nice longer walk.
no subject
Date: 2023-05-14 04:31 am (UTC)From:Exactly, how hard is it to at least to hold to that as an ideal?
I didn't even think about it as an abuse tactic in ship wars. Considering that Untamed ship wars is how AO3 got banned in an a small area of the world called all of China and Russia, woof, time to pour one out for the Untamed tag wranglers. Wouldn't want to be in their workgroup right now.
What tools does Ao3 even have to deal with spam except for locking down invites? They do not have enough people to hand-approve fic if they start dealing with being overwhelmed like a lot of other places area.
This is going to make discoverability for authors way worse than it already is.
no subject
Date: 2023-05-15 03:42 am (UTC)From:Even if it's difficult to moderate out of existence, having a STANCE opposed to it would at least help. Extending the rules regarding plagiarism to also include AI-generated fic would mean that it COULD be enforced against the same way everything else is - via user reports and investigation. They may need to expand their volunteer base again, but even so.
In general, I am glad that AO3 doesn't bow to userbase pressure - like anti campaigns to ban "problematic" content - when it comes to TOS changes. For the most part that is a good thing, and it's a necessary one to serve the mission that AO3 was originally created in service of. However, the issue of allowing AI "fic" feels antithetical to that mission, so this really does feel like a different situation to me.
And yeah, I foresee lots of potential for spam and indirect harassment. Considering how vicious various shipwars and anti crusades can be, I imagine it could get very nasty.
I'm not sure there's much spam mitigation outside of locking or significantly slowing invites. The site being invite-only certainly helps to a degree... but someone dedicated can and will just keep jumping into the invite queue with different emails to store up accounts to use, should some get swatted for spamming.
Discoverability is already so damn difficult. It's morally upsetting to me that actual writers can potentially be shoved down under a mountain of AI spam.
I'd *love* to believe that it just won't ever be that much of an issue, that the people who do use AI to generate text will all mark it clearly, that no one will use it maliciously, etc. but I haven't been that naive in a long damn time.
no subject
Date: 2023-05-15 04:52 am (UTC)From:This is one of the things that drives me nuts about the 'ai' debate. No moderation of disallowed content is ever 100%, do you things to make people think twice, disincentivize, etc. This is just how all content moderation works.
It's like people are constantly trying to hold moderation to an impossible standard or refusing to accept that people act like people unless you can create a logical framework for why a person acts like a person. I can't explain every aspect of the chilling effect involved, some should be fucking obvious and others just... are.
Also, people act like people and not like logical models of reason example #962: "I find it absolutely fucking baffling in the most hysterical way that we appear to have meaningfully shifted our spam account creation percentage downward by a permanent 5 points or so just by me amending our signup to tell people spam is against the ToS" link
I do not have it in me to explain basic norms of human behavior to people who think humans run on logic and don't have an emotional response when they post content online and it goes badly, even if the content is not theirs and the name is 'only' a psued.
Just... if people can only use plagiarism bots on the downlow then the spread will be a drastically less than if people are openly talking about it. This is how humans operate.
no subject
Date: 2023-05-16 06:06 am (UTC)From:Not everything has to be all or nothing! You can admit that it's difficult to 100% effectively prevent any AI work from being posted, while still making it clear that you do not WANT AI content to be present. Disincentivising does work, even if it doesn't work 100% of the time.
Like with the "we just told people not to spam, and it actually cut down on spam" thing. That seems super silly, but at the same time, it DOES make it unavoidable for the would-be spammer to know that they're breaking the TOS, which means they know they'll likely get banned should they still do so. It can create a perception that the site is more serious about spam accounts, and make it seem more likely that they'll get caught.
People keep arguing that "well, at least now people will tag AI stuff; they won't mark it if it's banned". And I get it; it's a common argument about certain content - "you want xyz content to not exist, but at least the xyz writers are tagging it; if you harass people for tagging it, or ban the tag, you'll just have xyz content that isn't clearly marked!"
I don't think that's a particularly relevant argument, though. Again, AI isn't about content per se. And some people are tagging it, which I appreciate from the "let me avoid it" stance, but plenty of people almost certainly AREN'T tagging it. I think the amount of untagged AI content would STILL be less were it made explicitly unwelcome. Permitting it, and especially *encouraging* it, means that there's likely to be a lot more tagged AND untagged AI content.
(Plus, the people who are trying to pass AI off as their own completely original creation aren't going to be tagging it anyway.)
People are people, and they'll act like people, and while some people will always feel like rules don't apply to them, there's a very much non-zero percentage that do NOT want to go against the TOS of a site, or don't want to share things that are disallowed, even if they don't personally object to the things in question.
no subject
Date: 2023-05-16 07:52 am (UTC)From:The overwhelming majority of people in fandom not only don't write or art, they don't even comment. They are shy to even kudos. Fandom is mostly lurkers because even under disposable psueds most people just can't or wont take the emotional risk of interacting online.
And those spammers are being paid to spam and have probably been hitting DW for years now. It's all freelance SEO optimization.
Yeah, the chilling effect of having to hide it being AI is a lot and means other people aren't going 'hey this person did it so I will, too!' Norms and 'monkey see monkey do' influence forums and communities and everything more than the actual roles. People absorb norms from peers.
People gunna people and people are weird. I can't provide a logical theorum, I can only say what works in practice
no subject
Date: 2023-05-17 03:59 am (UTC)From:In general, it seems like people WANT there to be a direct, logical, easily traceable pathway for why someone does what they do, and then they get VERY UPSET when that isn't the case.
Yup, it probably would be much easier to deal with a lot of things if people were perfectly, logically predictable, but that has literally never been the case ever, and I don't know how to get people to realize that!
That's also a huge part of it - the peer pressure/fandom trend/ooh, I think I'll try that! effect.
I think we've all seen that happen with tropes, for good or bad. One person writes something that other people like, and suddenly a particular fandom, ship, or even fandom as a whole suddenly has a LOT of the same trope being written, because people enjoyed it and want to try their own version. Sometimes it's great, very "yay, two cakes", or sometimes it sucks because it's a thing that you DON'T want to see, or bad fanon that you hate, etc. Seeing more AI works is likely to make more people post more AI works.
no subject
Date: 2023-05-17 06:14 am (UTC)From:"Seeing more AI works is likely to make more people post more AI works." Yup, that is key. That is just how humans are
no subject
Date: 2023-05-19 03:37 am (UTC)From:no subject
Date: 2023-05-19 04:13 am (UTC)From:I usually manage to avoid the trolls and the conservatives but it's like... honestly what I am most tired of is people getting that pushing back and saying 'no' is actually super effective. There is a lot of BS we didn't adopt because people went 'lol no' and while this has more legit uses than NFT bs, companies like Sudowrite will fall apart if we don't accept them as okay.
no subject
Date: 2023-05-20 03:55 am (UTC)From:no subject
Date: 2023-05-14 05:36 am (UTC)From:Just tossing an alternative out there for people.
no subject
Date: 2023-05-14 06:18 am (UTC)From: