That's the point. You can't prove the violation if you aren't privy to what is used.
That's actually the beauty of AI. You can pick apart exactly how much of something it uses to create its output.
Should we allow for AI-generated fiction writing?
- Thread starter Sgt_G
- Start date
That's the point. You can't prove the violation if you aren't privy to what is used.
That's actually the beauty of AI. You can pick apart exactly how much of something it uses to create its output.
There is no violation unless the work violates copyright. That's the point. What was fed into an AI is entirely irrelevant.
What it produces is relevant. If what it produces has violated your copyright, that is settled law and by all means pursue action.
The process by which the work was produced has absolutely no bearing on copyright whatsoever. It's an absolutely nonsensical argument.
Let me get this straight...
1) AI company uses copyrighted work.
2) AI company refuses to disclose sources.
3) Individual whose copyright was infringed and believes their copyright was infringed can't prove it because of #2.
4) Sucks to be that individual?
Is that really your line of thinking here? If not, what's your proposed remedy?
You're suing based on a similar product. How the product got created really isn't germane to the argument. The published text is what matters. How similar is it to another published work.
Actually, there are MANY copyright suits that deal with how the product got created.
Bones: Damit Jim, you can't sue a damn computer!
Spock: Logically Doc one could sue another at any time
Scotty: Aye but winning the suit is an entirely different matter.
Kirk: Who said anything about needing to win, the action in itself becomes the means to an end.
Scotty: Aye! A wonderful game of chess!
Bones: sounds like a back room game of poker to me.
Spock: Indeed.
Rand: (handing Kirk a deck of cards) The deck of Fizbin you asked for, Captain.
Spock: Logically Doc one could sue another at any time
Scotty: Aye but winning the suit is an entirely different matter.
Kirk: Who said anything about needing to win, the action in itself becomes the means to an end.
Scotty: Aye! A wonderful game of chess!
Bones: sounds like a back room game of poker to me.
Spock: Indeed.
Rand: (handing Kirk a deck of cards) The deck of Fizbin you asked for, Captain.
The AI company "using" copyrighted work is quite literally no different than a human reading a copyrighted work. The ONLY thing the AI company is on the hook for is to ensure they legally acquired the work. If they pirated a novel and fed into the AI, that's wrong for sure. But if they obtained it legally, it's been compensated.
Just like a human. If I acquire a work and I read it, I have been "trained" on that now. Do I have to continue to pay a content creator because it that work is now part of my knowledge base? If I were to publish a work, I now know money to every content creator of every single piece of media I have ever consumed? Do you see how absolutely ridiculous that is?
Now it only matters what occurs with the final product. If the final product violates copyright, there you go.
False. It's not HOW it was made. It was that the final product infringed on copyrights.
Vanilla Ice sampled "Under Pressure", leaving the musical arrangement the same. That violated copyright.
Had Vanilla Ice sampled "Under Pressure", but had edited the composition enough to that it was not recognizable as "Under Pressure"... there's likely no copyright violation.
It had absolutely nothing to do with the process that used to create the song, it was that the end product violated copyright.
In the absolute most simple terms, you generally don't have a copyright on the words, you have a copyright on the unique creative idea. (I say generally because you could potentially hold copyright on proper nouns, character names, etc.)
If I type out "A man from Texas bought a machine gun and then travelled an alternate dimension to fight space lizards but in the end he found that he was actually a space lizard himself and join with the space lizards to attack Texas"... I have a copyright on this idea. I don't own the words. I own the story.
If somebody were to read that (AI or Human, doesn't matter), and then reconstitute the words into something else... "A lizard man from the end dimension found out that he himself was a space machine gun, actually attacking space Texas to join with the travelling alternate lizards."
My second story didn't violate my original copyrighted story. It's a different idea. It uses the same words, which are NOT copyright, to produce a different idea.
And how can we know whether the AI company acquired the work legally when they refuse to release their sources?
This is a much more reasonable discussion I think to have, and where I can get on board.
I can see both sides here. I can see how an AI company would still be hesistant to disclose it's sources, fearing that the argument of "We just want to make sure it was obtained legally" will immediately return back to "Ok now pay me for literally anything the AI produces ever for the rest of time because it might possibly at some point use a word or two that I had typed", making the whole thing entirely unreasonable.
However I also don't think there is any real legal impetus on the side of the AI company to disclose it's sources. If I, as a human, write a novel... must I disclose what media I have consumed that may have potentially inspired my work? If the answer is "No", then... you have the answer as it pertains to AI too.
My litmus test for the whole thing is replace "AI" with "Person" and see if the argument is still reasonable. If it is not, the argument is not.
The thing is, copyright law isn't objective. It's incumbent on the parties who feel their works have been infringed to take action. If you have the resources of Paramount or Google or Microsoft or are part of a major trade organization, this is easy. If you're an individual who wrote a book once, this is much more challenging, and yet they should have equal protection.
If Paramount felt a random AI generator was using their media as a source, they have the resources to pursue it. If I feel the AI generator is using my sources, I don't have the resources to pursue it. So, what then...sucks to be me?
No copyright infringement claim is going to go to the holder on the basis of "a word or two that I had typed", and if you're seriously arguing that then I don't feel you're arguing in good-faith. However, if the AI is scraping my book without my consent, then I should have the right to know that it's doing so and be able to act to prevent it from doing so. It doesn't matter how much of my book the AI is scraping; it shouldn't have any access to my book.
On the other hand, if a human is being inspired by my book, then you're right; it's a lot lower-profile and I may never notice and find out...but if I do find out, then I can examine the book myself and compare it to my own work. But AI is churning out hundreds or thousands of works a day, while human creators are significantly less prodigious.
Paramount would be within their rights to issue a C&D to everyone who publishes fanfiction...but they don't because it would be more trouble than it's worth and it would look like they were just bullying their own fans. But if they thought an AI generator was utilizing their work to regularly produce "new" things, and especially if they thought those things were eating away at their bottom line? You can bet they'd take action.
If Paramount felt a random AI generator was using their media as a source, they have the resources to pursue it. If I feel the AI generator is using my sources, I don't have the resources to pursue it. So, what then...sucks to be me?
No copyright infringement claim is going to go to the holder on the basis of "a word or two that I had typed", and if you're seriously arguing that then I don't feel you're arguing in good-faith. However, if the AI is scraping my book without my consent, then I should have the right to know that it's doing so and be able to act to prevent it from doing so. It doesn't matter how much of my book the AI is scraping; it shouldn't have any access to my book.
On the other hand, if a human is being inspired by my book, then you're right; it's a lot lower-profile and I may never notice and find out...but if I do find out, then I can examine the book myself and compare it to my own work. But AI is churning out hundreds or thousands of works a day, while human creators are significantly less prodigious.
Paramount would be within their rights to issue a C&D to everyone who publishes fanfiction...but they don't because it would be more trouble than it's worth and it would look like they were just bullying their own fans. But if they thought an AI generator was utilizing their work to regularly produce "new" things, and especially if they thought those things were eating away at their bottom line? You can bet they'd take action.
So then, no human should have access to your book. It literally makes no difference.
That's fair, but I don't really see how it changes anything.
So ok, fine. The New Big AI discloses all of it's sources, your book is among them.
New Big AI generates 90,000 works a day.
Are you going to have the resources to go through every single on the of the 90,000 works in order to compare and see if any of them may meet the threshold of violating a copyright you own?
Or i'll even toss a bone here. The AI company develops a way that will not only disclose what works it has fed into the AI, but upon every single work generated the AI will also provide a word-by-word breakdown of what source it used (which is... absolutely ridiculous and is just not in any way reasonable or feasible), how do you then break it down? And the issue still remains that you will still go have to go through 90,000 works created per day to scan to see what percentage of your your words the work used.
AND THEN, just because it used your work as a source for say, 4.3% of the new work generated, now you need to read the work and compare it to see if it was actually a copyright violation or it literally just used words from your work. 90,000 times. Per day.
It would be absolutely ridiculous to demand that you are paid by the percentage that the AI detects in has used from a source, because even if it used 100% of the words from your work... that doesn't actually mean it has violated any copyright. It just means it used the same words you did.
Anyway you package it, it either doesn't make any sense or is just entirely not feasible and not workable. The BEST thing you can possibly get is for AI companies to disclose their sources so that you can potentially ensure the source was legally obtained.
My serious question... when it comes to compensation, as an author, what do you want? Specifically? What do you want to be compensated on, how would the compensation work? I'm genuinely curious because I haven't actually seen any true, specific demand other than "pay me".
EDIT -
From a purely pragmatic point of view, the "try and stop us people" once again do have the most convincing argument.
For the sake of argument, let's say there is a universal law passed, AI devs must disclose what works the AI was trained on.
Ok. Law is done.
AI dev doesn't include your work. You suspect your work was used. Prove it. From a US perspective, if you're trying to press criminal charges, prove it beyond a reasonable doubt. Civil charges, prove it 51%. Either way, prove your work was used.
I'm fairly certain it's impossible.
I'm tired of having this debate.
AI generators should not be allowed to scrape any works without the consent of the owners of the copyrights of those works.
AI-generated media is not equivalent to human-created media. AI-generated media at its best remains an elaborate copy-paste, and it's being done without even the most minimal attempts to credit the copyright owners, while humans who are publishing works that they're consciously aware are inspired by other works will usually undertake some effort to give those creators credit where it's due. Discovery is a new work of fiction, yet it still notes that it's based on Star Trek created by Gene Roddenberry. When I see AI-generated media providing credit to the sources it's pulling from, I'll reconsider my perspective.
Feel free to disagree, but I'm done arguing about it. If you want to consider that a victory regarding the strength of your arguments, then by all means, give yourself a pat on the back.
AI generators should not be allowed to scrape any works without the consent of the owners of the copyrights of those works.
AI-generated media is not equivalent to human-created media. AI-generated media at its best remains an elaborate copy-paste, and it's being done without even the most minimal attempts to credit the copyright owners, while humans who are publishing works that they're consciously aware are inspired by other works will usually undertake some effort to give those creators credit where it's due. Discovery is a new work of fiction, yet it still notes that it's based on Star Trek created by Gene Roddenberry. When I see AI-generated media providing credit to the sources it's pulling from, I'll reconsider my perspective.
Feel free to disagree, but I'm done arguing about it. If you want to consider that a victory regarding the strength of your arguments, then by all means, give yourself a pat on the back.
I just still haven't seen any reasonable argument as to WHY. It's just "it's mine they can't".
On the flip, The Orville is a new work of fiction, clearly based on Star Trek but changing things so as to not be in copyright violation of Star Trek and does not credit the source it's pulling from...
I'll take it as a mild win sure, if only because there has been little attempt to actually retort to my arguments other than "no they shouldn't do that." But also dropping a go nowhere circular argument is entirely acceptable. You clearly are not going to change your mind, I have been given little in the way of any sort of compelling argument to change mine.
I'll leave it at agree to disagree. And then go generate some AI stories.
Yes, because it adapts the protectable content (proper names and specific designs) of the Star Trek IP that came before it.
Clearly inspired by the unprotectable concepts of Star Trek (starships, exploration, inter-species space government) and therefore not "based on Star Trek" in a legal sense.
Exactly my point.
Exactly so. Those authors who have copyright have put in the time, effort, and research to create the content. There is a reason why copyright exists and that AI blatantly taking from it is considered "A-OK" strikes me as incredibly shortsighted on what it can do to future producers, or works not currently in public domain.
Indeed, yes. Put some damn rules in place over AI, rather than the blind, "Whelp, it's coming; shame you can't do anything about."
The insensitivity towards fellow humans in the name of AI is galling.
An AI is not a person.
Would somebody please post some AI generated fanfic to this channel so we can ban it already?
It's more along the lines of, if I buy a book, I don't have a right to make a movie out of it.
That's why there is a difference between who has the rights to the book, and who has an actual copy of the book.
The court case is to determine if AI usage violates the usage rights. And you can argue that it's no different than a human. But that, quite frankly, is pretty much what the court cases are going to determine.
Last edited:
Similar threads
