I keep having to repeat until I'm hoarse: "mere profit potential does not confer legitimacy on a business model"—see also child pornography and illegal drugs: both are profitable, neither of them are legal.
If your business model cannot operate without copyright violation then it should be no more legal than any other large-scale IP piracy operation.
@cstross@lilithsaintcrow There's a big difference between copyright violation and fair use. Many entire industries rely on fair use of copyrighted materials.
@cstross Given capitalism's inevitable rapaciousness, it's no wonder a lot of folk are (consciously or otherwise) conflating large-scale grift with "business". It's the Gilded Age all over again, and we all know how THAT worked out.
@lilithsaintcrow@cstross This is what happens when humanity didn't develop the "hitting the monkey who tries to grab the hanging banana" collective behaviour whenever some fucko goes "I have discovered a new way to become extremely rich extremely fast".
@Illuminatus I am amazed, every ding-dang day, that the guillotines have not come out. It appears humanity is, on the whole, far more patient and forgiving than I ever thought possible.
Not quite sure what to do with this, since it flies in the face of my long-held cherished certainties, but here we are.
@cstross@lilithsaintcrow I do not think they said that they violated copyright, but that a language corpus of the required size necessarily includes copyrighted material.
@lilithsaintcrow@cstross a) they wouldn’t have to, if it were fair use b) they cannot, given the size of the corpus and the sheer number of copyright holders c) not they (i.e. the researchers and startups doing the hundreds of other language models), cannot either. I admit though, that b) and c) are moot if not a), and then not only OpenAI would be DoA.
@fpbhb@lilithsaintcrow Fair use (or fair dealing, in the UK and other jurisdictions) has a tight legal definition and indiscriminately slurping an entire author's corpus into software and then regurgitating any or all of it on request is not fair use.
Never mind OpenAI having admitted to using content slurped off the internet that included piracy websites with illegal copies of thousands of books, which OpenAI didn't bother to check for or sanitize.
@cstross@fpbhb@lilithsaintcrow reading it all may well not be an offence (there are good arguments for that - like making internet search possible). Copyright requires an act covered by copyright is performed - clearly "spewing forth a copy of a copyright work" is covered. Probably distributing the model does if it can do so, but reading the internet is no different to a web indexer, or scanning for criminal material
Absolutists might want to be very careful what they wish for.
@etchedpixels Some German (at least) news publishers have already argued that (and gotten into compensation deals) because the snippets displayed during search were already so good that people didn't follow the links anymore.
But there's a difference between indexing (and returning results that clearly link to the original source) and remixing/reproducing/obfuscating said source(s).
@larsmb@etchedpixels@cstross@lilithsaintcrow It is noteworthy though, that what the search engines did actually was allowable by German copyright law, and new law (“Leistungsschutzrecht”) had to be invented to funnel some money from Google to Springer. Whether that was good or bad.
@fpbhb@larsmb@etchedpixels@lilithsaintcrow I'm not sure who's worse: Google or Springer. (Both of whom are pretty ghastly—Google for ditching "don't be evil" and enabling the entire SEO-riddled behavioural advertising web, and Springer for being, well, Springer, whose motto ought to read "DO be evil".)
@cstross Pretty very fucking sure Axel Springer is by far the worse of the two.
Coincidentally to this discussion, they've just announced a partnership with OpenAI for content creation.
@fpbhb@larsmb@cstross@lilithsaintcrow Agreed. And there is a legitimate debate to be had about whether there needs to be some kind of collecting society arrangement for some classes of compulsory licensing of works for certain AI uses. I'm on the no side. It's one thing for an AI to eat the entire Stross and be able to summarize him as an author and another to say "you have no choice but to let our AI write stories like yours but you'll get 2p a book"
@lilithsaintcrow@cstross@fpbhb That's what search engines do. It's what humans do (but for smaller subsets). Nobody stole your work by indexing it any more than someone did by reviewing it or you stole by reading lots of articles about how to be a successful writer and then writing and selling a book.
If you want a world where your book royalties are then sliced up with everyone whose articles you've ever read on how to be an author then carry on - that's where your model leads.
@etchedpixels@lilithsaintcrow@fpbhb That's actually a good remuneration model—microbilling along the lines of Ted Nelson's Project Xanadu hypermedia from the 70s and 80s!
But really, indexing isn't the problem: it's unauthorized reproduction that violates the author's copyright.
@lilithsaintcrow@cstross@fpbhb If you have nothing better to do that wilfully misinterpret me you know where the block button is.
I'm old enough to remember the madness of the napster era where the music industry wanted anyone who so much as accidentally accessed an internet site allegedly containing pirate material to get their internet cut off.
And I've most certainly never said that intentionally pirating a copy of a work is ok.
@etchedpixels@lilithsaintcrow@fpbhb The RIAA back then was targeting end users for piracy. Going after OpenAI Inc is more like targeting Napster directly, as in the company making money off the p2p piracy software and not compensating the creators at all.
@cstross@fpbhb@lilithsaintcrow On pirated web sites - if they can't do that, then neither can the people who look for pirated content on the web to take it down, because unless they had rights to every work in the universe they might access a work they didn't have access to. It would also imply the kind of madness the music industry wanted where they could prosecute people for merely accidentally accessing a pirate site.
More of a problem is stuff like CSAM and personal data.
@larsmb@lilithsaintcrow@cstross Thank you. That is pretty much my point. Is it legal and ethical? If no, fine. Then it can’t be done. But all this “slurping”, “theft” and “eat the rich” rethoric is less than helpful.
@fpbhb@lilithsaintcrow I'd be willing to countenance a limited exception for research purposes only, on condition there is NO commercial reuse of the resulting LLMs unless all rightsholders have received financial compensation. But startups trying to move fast and break [other peoples'] things should be straight-up illegal.
@fpbhb@lilithsaintcrow@cstross b) is just "their business will fail if they have to pay for licenses for the IP that they're using". That's not a reason why they shouldn't pay, it's a reason that their business model is broken and should be abandoned.
Failing to make big businesses pay for externalities because they'll become unprofitable if they're made to do so has already caused so much damage in the world (e.g. climate change), we really should have learned to stop doing it by now.
The profitability argument was previously used to defend child labor, and slavery before that.
I'm comfortable saying if your business depends on slavery, child labor, pollution, or theft of the work of others... then your business is bad and should not exist. You can learn a better business model.
@fpbhb
> a) [LLM developers] wouldn’t have to [pay for copyright], if it were fair use b) they cannot, given the size of the corpus and the sheer number of copyright holders
@pluralistic has written some good stuff on this topic. There are podcast versions here that link to the texts:
@fpbhb
> a) [LLM developers] wouldn’t have to [pay for copyright], if it were fair use b) they cannot, given the size of the corpus and the sheer number of copyright holders
@pluralistic has written some good stuff on this topic. There are podcast versions here that link to the texts:
@fpbhb@lilithsaintcrow What part of "a language corpus of the required size necessarily includes copyrighted material" do you not recognize as being copyright violation?
(It's like being "a little bit" pregnant: it's an all-or-nothing condition, no grey areas apply in this context.)
@cstross@lilithsaintcrow just because something uses copyright doesn't make it violative of copyright.
I think generative AI is so distasteful that it's turning people into copyright maximalists without them realizing that's what's happening. Is there evidence that llm makers are using copyrights in ways that are actual violations? (I legitimately don't know the answer to this, and of course if the answer is a clear yes, that matters a lot here.)
@Amoshias@lilithsaintcrow The answer to your reply-guy question is "yes". Go do your own research: there are class action lawsuits in progress over this shit.
Jesus Christ. Is that the new thing to avoid having to actually engage in discussion or think in any way about the points you are making? If anyone questions your viewpoint in even the tiniest way they are a 'reply guy'?
I'm sorry that "I agree with you 95% but there's a serious issue to consider" isn't pure enough for you. Have a good day.
@cstross@lilithsaintcrow please find me saying piracy is good for authors, here or anywhere. Ever.
You WILL find me saying that copyright maximalism isn't good for authors, or anyone but megacorps. You will find me saying this all the time. It's not even in the same GALAXY as "piracy is good for authors."
@Amoshias@lilithsaintcrow I have no idea who you are; you have no visible Masto profile and a total of 65 toots since you joined an instance I've never heard of 32 days ago. In twitter terms of yore, you're an egg.
@Mschatelaine@cstross@lilithsaintcrow 1. This wiki page is an essay, not an article on the state of the real world, as it clearly states...
2. It also clearly states "The copyright status of LLMs trained on copyrighted material is not yet fully understood."
@Amoshias@Mschatelaine@lilithsaintcrow Wikipedia is not authoritative on rapidly-moving current events (with the lawsuits around LLMs are). You need to go to primary sources.
@Wikisteff@Amoshias@lilithsaintcrow again with the OP - they admit that they can't run their business model without violating copyright, and there are lists of artists and writers whose copyright has been actually violated so your whataboutery is moot
@Mschatelaine@Wikisteff@Amoshias@lilithsaintcrow They absolutely did not say that. They said they can't run their business without fair use of copyrighted materials. Just like the movie reviews industry can't either.
They claim their "AI" is "trained" on other people's art and writing.
By saying "training", they're trying to evoke a cybernetic schoolchild, sued by Tolkien's estate for reading the assigned book, The Hobbit.*
But their "AI" is nothing but spicy autocorrect, with the full text of The Hobbit used to match data and spew out pseudo-Tolkien "Art" at the press of a button.
*Example used after a recent copyright case. No hobbitses were harmed in the writing of this.
@davidtheeviloverlord@cstross@lilithsaintcrow "Training" is a long-held technical term in computer science. The term is useful within the practice and not intended to evoke anything from the layman. You might as well complain about the technical term "binary tree" evoking ideas about gender roles in plants.
@cstross@lilithsaintcrow I fear that that’s a narrative we lost control of when we allowed online platforms and search engines to enclose online tracking data for commercial exploitation. Since they got away with that in the name of “innovation” everything that can be freely accessed (i.e. without technical restrictions) is fair game in developers and executives eyes.
@cstross@lilithsaintcrow The scary thing is how many critics of copyright over-protection (which is of course a thing, as we all know) are now seeing generative AI as the thing that will break the copyright stranglehold, without realising that it is merely another form of enclosure. Just this time not one that bestows any benefits on the original creators.
@cstross@lilithsaintcrow Generative AI does not release over-protected copyrighted material into the public domain as some claim. It merely uses it to line someone else’s pockets in the same way as Big Tech harvesting the “behavioural surplus”, as Zuboff calls it, rarely benefits anyone but Big Tech.
@nuncio@lilithsaintcrow Some of us have been onto the grift almost from the beginning. LLMs is where the grifters who'd been suckering mugs out of their savings using cryptocurrency scams shoved their money when the blockchain bubble burst. And the original dot-com wave before that. There's a new bubble every decade these days.
You are FAR from the only one pissed off about this.
Nobody can call me unreasonable, though. My store offers a convenient AI-friendly license. They can use documents derived from all my independently published work for only three million dollars a year.
@mwl Yeah, if they offered me a proper price commensurate to their own profit, I’d consider the notion. A few million would be acceptable for one or two books' worth of training material.
@lilithsaintcrow@mwl@lizmonster as far as I know, they're not really making any profit. LLMs are crazy expensive to train and run, all the current hype only exists by venture capitalist money being burned. And theft, of course.
@sophieschmieg
A lot of this discussion is based on the idea that LLM are just techbros full of money, while writers are poor and starving. The objective is to kill this revolution like they killed Napster: big tech and big content merge (the Washington Post is owned by big tech already), leaving free software and little creatives in the rain. #ai @lilithsaintcrow@mwl@lizmonster
@lilithsaintcrow@gelato_al_pollo@sophieschmieg@mwl Also, AFAIK (and I may be wrong), research into actual artificial intelligence is a whole different branch of science than LLMs/Midjourney. So if we're talking about "the AI revolution," LLMs have nothing to do with it.
All of these "generative AI" companies have always been welcome to contact me, ask my permission, and offer me appropriate financial remuneration. Hell will freeze over, of course.
Does the American law grant an author exclusive rights to train LLMs? European law, if I understand it correctly, expressly says that such rights do not exist and can not be sold, but it might be a local thing
@mwl@lilithsaintcrow@lizmonster Oh yeah, agreed. Like someone else said, they're burning through venture capital. Lack of profit doesn't mean the CEO's aren't taking home offensive salaries. I take that to mean, by the way, they could go the way of NFT's any minute now. Doctorow thinks it's a bubble, and I think he's right. At a certain point, if it doesn't make money, they'll stop, and it's proven unable to replace labour, which was the goal.
@orionkidder@mwl@lilithsaintcrow@lizmonster
Nail on the head. The point of crypto was to evade taxes and/or financial regulation, and the point of AI is to avoid paying humans for labour, and these vampires will keep on looking for the magic cheat code to make capitalism invulnerable to its own defects.
@lilithsaintcrow@mwl@lizmonster Absolutely agreed. Whether it works or not, the threat of it can be used as a cudgel. They don't actually care about automation. They care about reducing costs to zero and keeping all proceeds. They'd be perfectly happy with slavery if it were an option.
@lilithsaintcrow@mwl@lizmonster My experience includes being in the negotiating team for a union, and the people we were across from at that table, they a) took it as their job to keep us from getting anything at all or preferably to push us back a worse deal, had vaguely no knowledge of good our job was done, and did not respond to reason or logic, like at all...
@lilithsaintcrow@mwl@lizmonster And b) seemed to truly believe that the point of labour was to shut the fuck up and take whatever the employer feels like giving based on how much the executives want to get paid.
So my answer is, it's ideology fuelled by greed. Greedeology, if you will. It's not a mental illness. It's not Satan. It's a belief system that they rationalize bc it supports them living in luxury and "getting to" treat people like shit.
@lilithsaintcrow billionaires: “we have to steal to buy more yachts you see”
…
Poor person “I have to shop lift baby food”
.
Guess who goes to jail
🤷♂️
@lilithsaintcrow
When you find out how much water and power this AI uses, it is madness to even pretend that this has any benefit or value to society. Just like crypto and NFTs, it sure feels like a billionaire scam to me.
@dsfgs@lilithsaintcrow
Crypto is used to pay for phony worthless NFTs. Bored Ape anyone? What a sick scam. I actually remember seeing celebs on the chat shows push that crap.
In 2018, they pushed "bitcoin will destroy planet by 2020", FUD-merchantry, and false. Bitcoin is infinitely better for planet, reduces consumptionist ideology and more.
@lilithsaintcrow@lin11c
The fact is that #bitcoin is the only censorship-resistant money, Wikileaks had to use it to report on war crimes when the legacy platforms like visa and (DARPA-funded and TwitBuyer-produced) Paypal blocked Julian Assange and Wikileaks.
With Lightning Network (LN) (Layer 2 bitcoin) infinite transactions are possible for one high energy transaction. The more people federating in the LN the easier bitcoin will be to use for artists, sexWorkers and everyone.
@lilithsaintcrow Archive.org and Gutenberg project have millions of pages of uncopyrighted prose. If AI trained on that, who knows, it might lead humans back to using apostrophes correctly or knowing the difference between the verbs "to lie" and "to lay."
@lilithsaintcrow I know even billionaires/corporations are not a monolithic class, but it's heckin ironic how throughout the 90s/early 2000s there was the whole weepy "piracy is stealing!!111" BS from media corpos, but now that other corpos are doing piracy, they're all like "uwu fwee speech we cannot do it withwout the copywwited matewial!"
@lilithsaintcrow@epidiah “Limiting training data to public domain books and drawings created more than a century ago might yield an interesting experiment, but would not provide AI systems that meet the needs of today’s citizens.”
@lilithsaintcrow I agree and I dont.
All human achievements are built on the achievements of others. I have a slightly hostile opinion towards IP laws. I can see some value in them, but largely feel the current rules do not provide the value they promise.
Also I'm all for AI not being exploitative to creators.
@lilithsaintcrow and yet the cowards won't join the copyright abolitionists. must me the delicious Microsoft dollars. or just the good old propagada of monopolies being great for artists!
@lilithsaintcrow well, i could speculate why you seem to think copyright is a good idea. but for the chance of a fruitful discussion, i suggest you explain it yourself.
or i could explain my position. but not right now, i really have to go to bed 😅.
@lilithsaintcrow @briankrebs
This thread is just delicious. I thought it would be straight outrage to enshittification in the responses, but the guillotines and manatees... Thank you. 😋
@lilithsaintcrow “Because copyright today covers virtually every sort of human expression [...]". So they finally agree: today's copyright framework is A Bad Thing.
(Copyright had sense when it's length was proportionate)
@lilithsaintcrow
What are your ideas on software? Should you pay Mastodon developers the same way the LLM companies should pay you? After all, software engineering is intellectual work, and you are profiting from it. @mau
@lilithsaintcrow i am not sure what you mean with "for decades if not centuries". Do you really believe that my artwork should be protected for decades if not centuries, or you are just saying "since some centuries ago [let's say from the beginning of XIX century] artists are struggling foi being fairly recompensed?"?
@lilithsaintcrow question: if I read your books, become "inspired" and write something similar - would I have to pay as well - besides paying for the original book?
True. It's shameful how many beautiful business plans are shot down by greedly little men wanting money for their goods or services. Do they not realize they could help make such truly beautiful profit for me and my stock-holders? Have they no soul?
Essentially, this is how Sean Parker and a few other assholes like Kim Dot Kim or whatever name that useless ugly fuck was using, got rich.
Theft of intellectual property. Napster and the like. Shut down and sued, eventually.
As I recall, the RIAA was suing individual users for like 300K during the process before they were shut down. Something like that seems to be appropro here and is probably already in the wings, along with suing these fucking grifters at the top.
They’ll get everyone. Maybe not the big players like Microsoft because at some point they’ll see the writing on the wall and bail. But the new generation of assholes like Sean Parker, absolutely.
Add comment