(no subject)

[olivermoss]

* Here's hoping the heat is over. Every time I blink the ten day forecast changes drastically.

* Last night the forecast was either for massive thunderstorms and flash floods or... none of that. The Sky went grey for a bit. Had some lightening and isolated showers, but the sky never broke into the big storm that was possible.

* IANAL but, if the OTW is open to works that cannot be copyrighted, how can they defend them against plagiarism or commercial exploitation?

We envision a future in which all fannish works are recognized as legal and transformative and are accepted as a legitimate creative activity. We are proactive and innovative in protecting and defending our work from commercial exploitation and legal challenge.

It has always been the stance of the OTW that fan creations are theirs, even if they don't have the rights to the world and the names... the transformative work is theirs. But, if people are posting bot output... if works with no transformativeness because there is no human authorship are put in the same bucket as works they defend, doesn't this weaken AO3's stance and ability to enforce things?

There is no current case law on specifically text-based bot output... but I think the guidance of the copyright office and the copyright office's ruling to deny copyright to midjourney art used for a comic makes the current stance pretty clear.

(People keep bringing up this the lack of specifically case law on specifically text based works and I think they are missing that the copyright office has been consistent on human authorship being key since even before the so called 'ai boom')

"We conclude that Ms. Kashtanova is the author of the Work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements," reads the copyright letter. "That authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the Midjourney technology are not the product of human authorship."

Basically, the writing and the putting of the images in that order was done by a human and that gets copyright, but not the individual images themselves. And this... makes sense to me. If I write a story with long passages of Shakespeare in between scenes because I am weaving the story of Othello and a coffee shop meet cute together to create something new... I have copyright over what is mine, but not over the Othello passages.

If we put work that has no proper authorship, no copyright, in the same bucket as works the AO3 claims to protect the rights of... isn't that a big fucking problem? Especially if AO3 doesn't know which is which? What happens when AO3 goes to bat for a work that turns out to not be transformative? If Ao3 was mislead about the work that is one thing, but if they openly invite computer generated content and treat it the same as human stuff, that's a problem.

Some people claim that fanworks are on rocky legal ground, if that is true, doesn't this kick away some of the rocks because suddenly AO3 doesn't care if there is a transformative element? And yes, only humans can make stuff that is transformative.

Comments

[mistressofmuses]

I hope the heatwave breaks and stays broken!

See, this stance on what works are copyrightable makes more sense to me - the idea that AI-generated content is not subject to copyright, and therefore attempting to protect it WEAKENS AO3's stance on transformative works, because so much of that transformative nature as previously understood hinges on the creativity of an actual person behind the work.

I keep hearing more people claiming the opposite, that OTW *has* to go to bat for AI-generated text, because otherwise it weakens the argument for transformative work as a whole. I think they're saying a machine "remixing" existing work scraped for a dataset is not legally different than a person writing a fic based on an existing canon, so allowing the latter but not the former weakens the stance of the latter. But I feel like I just don't get that? The human authorship aspect IS an important piece of the definition imo, and excluding non-human-authored work is NOT a horrific slippery slope to content bans.

I'm also not a lawyer, and I'm not going to claim to be. I don't have the expertise to say where things currently stand or where it's likely to go, or whether one of those arguments is the "correct" one from a legal standpoint. But, to me, I think human authorship should be an important distinguishing characteristic, and I don't think human-input-prompt satisfies that requirement.

I also keep seeing "ugh, people claiming AI is plagiarizing when you're writing fanfiction which is already stealing characters and settings anyway is soooo hypocritical", and it's so extremely frustrating to me. That whinge is mostly coming from people who ARE fic writers, and previously have been very clear about how legitimate they find fanfic... and yet somehow they seem perfectly willing to now say that it's ALL plagiarism actually.
Which when part of my frustration with AI as a whole is that I think it devalues creative work, welp, there is already is.

[olivermoss]

Yeah, there are weird shades of Geek Social Fallacy #1 in people thinking we haven't have any boundaries, because what if that effects me?? People are not bots. That people think these programs are anything like humans is very concerning.

Also, despite their maximally inclusive stance, they already delete a ton of stuff! The archive does have standards for what counts and what does, what is transformative enough and what isn't, etc. They just don't make their standards public or make a big deal about deletions. How do I know? many reasons, but let's look at ones I can cite! First off, can people post orig fic to the archive? From the FAQ 'Yes and no.' And then they go on a bit, but basically if your work feels fannish, is okay. If it don't, nuke. AO3 does allow original fiction but is not there to be a host for just anyone, only people coming out of fannish culture.

Citable thing 2: Look at the March Meta Matters comm on DW. The assumed posting a range of things to AO3 was fine and wound up with an account suspension. Some fannish non-fiction is okay, other stuff gets nuked.

AO3 is constantly deleting content and if people think they aren't... I don't even know what to say to that. That they keep it and their litmus tests on the down low keeps things calm and makes it harder to fuck with them.

Semi tangent: Sexy Times with Wangxian-gate. When people start going 'oh no tag spam is breaking the archive' people started uploading War and Peace to the tags in hopes of breaking the archive. Make the exact details of deletions, suspensions, litmus tenses, etc public then the people who want to fuck the archive will jump on it! Like they did with tag-gate. This is why I expected Ao3 to just not make a public statement, being kind of distant and monolithic keeps problems tamer and helps keep the abuse team from getting even more overwhelmed.

Semi tangent 2: When they made changes to gift-fic people were upset. How was that supposed to tame abuse? That was a massively prolific abuse vector, they just didn't say ahead of time 'well, by the numbers here's the best way to fuck with authors and drive them from fandom'

uh, anyway, tl:dr we have lots of barriers and rules already. Putting on rules for 'ai' isn't a slippery slope, we've always had lots and lots and lots of barriers and standards and rules.

And yeah, human authorship is key and people need to stop thinking bot output is anything like that. I promise you if a trained a LLM on just Stephen King... it wouldn't work, despite all the people claiming you can train on a single author, it's BS, but let's pretend it's not! If I trained on all on just King and was posting the results, that would clearly be crossing a line by any standard. But I can read a bunch of King and write fic in his style based on his ideas.

[mistressofmuses]

Right? That's very much how it feels - the idea that any rules are unfair gatekeeping. There are absolutely negative forms of gatekeeping, and unfortunately a lot of those are common in certain geek spaces... but then there's also the perception that all gatekeeping is bad, which is also not the case.

And exactly. AO3 is based on broad inclusion, but that doesn't mean there are NO standards or rules about what belongs and what doesn't.
It's largely a good thing that they don't speak publicly about what's allowed or what has been removed, because it does prevent a lot of drama and pressure.

As you said, not speaking publicly also does prevent deliberate abuse of certain features or exploitation of gaps in site functionality/stated rules. It's fine if people *don't* think of the ways in which something could be weaponized against the archive or against individual users, but then people are sometimes weird about the idea that something is implemented or changed in a way to prevent those abuses.

Bot output is not anything like actual human authorship. I really don't like how many people seem to claim they are - whether that's the outsider assuming that all fanfic is garbage anyway/writing is frivolous or easy/etc. and truly believing that their bot-generated shit will be an easy moneymaker if they flood submission forms with it. Or whether it's people "on the inside" of fic and writing communities who seem to think that readers are dumb enough not to tell the difference, and who will devalue their own and others' work in order to defend their right to potentially "produce" faster or easier.

[olivermoss]

Precisely. There are a list of reasons why the old gift fic system was a particularly good abuse vector, but most people simply would never think of it. Being too open with certain information is like drawing a map for bad actors. I am not saying them being non-transparent in general is good, but there are many reasons to keep a lot of stuff in regards to abuse / removal on the downlow. I don't even know how to have a convo on this topic with people who don't get that the archive is constantly removing content based on the content, not just spam or ko-fi related removals.

A bot can only seem like human authorship if it's taking an existing story and tweaking it slightly. Bots have 'accidentally' put in whole paragraphs of other people's writing unchanged. Bots aren't 'learning' they are more and better data to search and humans have spent more and more time on maintaining / organizing / labeling the data. People are like 'it doesn't have your data, it learned and moved on' No. I think it was in the Wired article I skimmed where they outright said Sudowrite people said they need the AO3 scrape or the program's functionality would be impaired. Yeah, it's trying to do plagiarism through obscurity, using the sheer amount of fic to hide that it's not just typical 'ai' levels of plagiarism, it's far more direct that that.

And now John Rogers is tweeting about know we know where the scrapes came from due to Omegaverse tropes and... I am having flashbacks to the day I found out that the thread that is credited by fanlore with the birth of omegaverse... some of the comments in that thread are me. I wish I never found that out

[mistressofmuses]

Yeah. There are benefits and drawbacks to transparency. Lots of people claim to (and maybe genuinely do) want complete transparency, yet oversharing can lead to a lot of people taking advantage of an exploit, or can lead to escalating drama. AO3 NOT being open about what they've removed and why is by and large a very good thing, imo. (And that's the case even when it allows shitty people to lie about moderation decisions that were made against them.)

And yeah... bots are more or less just... really advanced text suggestions. Like... if I'm sending a text, the app I'm using has been trained on my general texting patterns, as well as presumably other data sets that tell it what words likely go together. Sometimes it's pretty damn good at guessing exactly what I wanted to say, generating a coherent message... but that's usually because it's something I've said before. Other times, it "suggests" utter word salad. I know that's a very tiny and basic comparison as opposed to the various text-AIs that we're really talking about... yet those text AIs are a lot more similar to predictive texting than actual authorship.
(And yeah, with a lot of human curation going on behind the scenes.)

Re: the Omegaverse thread. Oh nooooo, lol. I'm so sorry. (Though to be honest, I *love* that omegaverse exists, because it's such a strange thing to have achieved such popularity. [Not that strange, maybe. People are horny weirdos, which I mean in an utterly complementary way.])
The other day I discovered one of our employees had a state-generated identifier that was ####ABO, and I laughed, and then could explain why to NO ONE.

[olivermoss]

Transparency sounds good, but it's complicated. There are some things AO3 should be more transparent about but not this. But it is frustrating that people think they aren't constantly making judgment calls about nuking content.

Omegaverse is fun and iddy. It's kind of interesting how it mutates and changes. I rarely find fic in it I like, but when it work, it works.

And yeah, finding things amusing and not being able to explain why or just deeply not wanting to is part of fandom life :)

[mistressofmuses]

Exactly. Transparency is one of those things that's used in a buzzwordy way as an always-good, always-needed thing, but there ARE actually drawbacks in some circumstances! Though there are things that would be served better by allowing more transparency than they do.

That's a lot of how I feel about it as a trope, too. Pregnancy is a major squick for me, which rules out a lot of the trope's content, but there've been a handful of stories within the trope? subgenre? that worked very well for me.

Lol, it very much is.