It’s no secret that Algorithmic Artificial Intelligence platforms vacuum up copyrighted materials and barf out an approximate imitation of human discourse. The rules on this board have always forbidden posting more than small bites of copyrighted materials. When using an AI chat bot, sometimes information is cited and sometimes sources have to be requested, but it all comes from somewhere else because AI doesn’t actually exist.
Should the same rules apply to posts quoting AI?
Cannot copyright random or unpredictable word salad.
Or do you mean the output of a search engine, where it comes up with an actual article? Or when it comes up with an article or link to an article, but that article is AI slop?
There are reasons to oppose use of AI, but that’s not one of them. If we prohibited all content from entities that learned language from consuming copyrighted material, we’d have no content at all.
So, the first issue is that we want human-generated content here, not AI posts. While it is ok to use AI as a starting point in a conversation or to lead a discussion down a possible path, you should not be posting large amounts of AI content. Along similar lines, an AI answer in FQ is not considered to be a factual response.
The second issue is that only humans can copyright something. So if an AI creates something, by law it cannot be copyrighted. The trick to this though is that it is possible for AI to create something that is effectively a “derivative work”. AI often lies about where it gets its information from (I have encountered numerous instances of the cite that an AI gives not actually giving you the information that the AI claimed).
Moderators don’t have the time, or in many cases don’t have the ability, to track down every place where an AI might get its information from. Asking us to verify whether an AI bot violated someone’s copyright isn’t practical from a moderation point of view.
I think it’s best to avoid AI posts as much as possible on the SDMB. In some cases it can be useful, but mostly it’s not something that we want here anyway.
ETA:
I should also point out that any AI generated content on the SDMB must be clearly labeled as such.
This gets blurry when AI is used to “polish” a paragraph or an image. Do I suddenly lose the ability to copyright my work if I let AI do some light editing?
The point I am trying to make is that if you create something and AI effectively makes a derivative work, that is still your copyrighted work.
But if an AI bot creates something that isn’t a derivative work, that cannot be copyrighted.
And yes, it does get blurry, especially when you can’t tell where AI got its info from.
First, what is the legal situation? The University of South Florida gives some guidance.
Material created by generative AI tools does not currently receive copyright protections in the United States. However, generative AI tools can create ‘new’ material that infringes on existing copyrights. Using this material in a project, presentation, or publication would have to abide by copyright law.
That leaves a gaping hole: how would anyone know whether a piece of text infringes or not? Many times we don’t even know whether text ostensibly from a poster in original or copied.
What normally happens here is that any very long amount of text that is copied from an outside source is removed by a mod unless the source is explicitly in the public domain. Shorter quotes are allowed through “fair use.” No one definition is available for what falls into fair use. Most short quotes that take only a small portion of a work qualify. (Technically, poetry and song lyrics were in the past banned from fair use because they can be so crucial to a small original, but in reality they are all over the internet. I’m a copyright sticker and I would use them online, but not in a printed book.)
It’s hard to imagine a situation in which someone would quote a very long AI-text that is entirely composed of copyrighted material in the first place. Nor would it stay up past the point where a mod notices it.
Posters are supposed to say outright whether a quote is from AI. For now that’s probably the best policy. I’ve noticed a marked increase in quality in summaries from Google AI when I already know a subject but am looking to answer a question succinctly. I still check the sources they cite to be sure. They still sometimes don’t directly back up the summary, and so I don’t use it. If I did then I’d hope that someone would correct it in a response just like any other unsupported assertion. The SDMB gives good feedback.
And just a reminder, we did have a community discussion previously about how we want to handle AI content on the board:
It’s not the exact same discussion as this one, of course, but there is quite a bit of overlap. Just in case people want to see what we talked about the last time we had a major discussion in ATMB about AI.
My understanding is that human input retains copyright. Most of what used to be called print text always was subject to polishing by editors, copyeditors, proofreaders, censors, and others down the line. Sometimes the work was almost totally rewritten. The original author was still given copyright, even if they had objected to the replacement words. Given that precedent I can’t see how an AI polish would be any different.
That should give people pause. You’re responsible for what you put out under your own name. Claiming after the fact that AI wrote something objectionable won’t get you very far, except in extraordinary cases.
Well, one big difference is that what a lot of people claim is “their own original work with just some AI polish” is about two sentences of prompt being converted to three pages of output.
Sometimes “The idea and story are mine” is actually “I typed in three names and a time period, then saw that the AI generated result is sellable product. Ka-CHING!!”
That would not pass muster, obviously, but, considering the amount of (human) work it takes to master the various tools (call them AI tools or not) and produce an actual saleable product, there is something to consider there.
ETA we have to distinguish the video game where you type in a handful of words and get a story or an image, from real creative work (often involving modifying/training your own AI tools not to mention knowing what you are doing in the first place)
At what point does “author” stop and “computer programmer” begin, though?
I have seen, for example, a musician work with and write actual, non-trivial computer code in order to produce experimental algorithmic music. She even mentioned that it was different from programming every single note and that sometimes the computer output would surprise her.
If you are working with this stuff, the artist may need to become a computer programmer because programming the computer is now an artistic technique.
We’re having an interesting discussion so far about copyright.
Which IMO wasn’t what the OP was talking about at all. IMO the OP meant this:
Which is of course utter nonsense.
Said another way, either @BlankSlate is confused about AI or I am confused about their post.
Yes it does. That is a decades-old industry term of art and has been applied to much, much less complex systems than those that exist today.
Correct. AI as it is portrayed in science fiction, where a fully sentient being pretty much indistinguishable from a human is created with computers, that does not exist. But that’s not what people mean by “AI”, any more than when you are talking about a “hover board” that kids play with, that you are talking about the anti-gravity board from the Back to the Future franchise. Just because a term becomes popularized in science fiction, that doesn’t mean that we must forever use that meaning for that term.
Anyway, back to the copyright concern, I think it’s fair to say that whether or not AI (as a whole) is guilty of copyright violations is at most speculation. Certainly, if it had been determined to be a copyright violation, then the AI industry would not be able to operate as it is. Unless and until it has been made illegal by the courts and/or new laws, it would be unreasonable to treat all AI as a prima facie copyright violation.
Or to put it even more simply, the answer to the question in the OP and thread title is an obvious “no”.
We’re not talking about all AI generated content from everyone anywhere anytime. The issue is whether quoting some AI answers on the Dope is a copyright violation.
Small bites of AI generated content shouldn’t be treated differently.
I was thinking insisting that peanut butter doesn’t exist because it doesn’t contain butter.
That’s another potential argument, sure. And similarly, that argument is ignoring the fact that the word “butter” has more than one meaning in English; it can refer to a churned product made from butterfat, or a spread made from pureed nuts or seeds. Just as the term “AI” can mean true intelligence in the form of a computer program (which does not exist yet) or the practical terminology that refers to the real-life industry of machines helping solve logic problems.
By the way, I had things a bit backwards; the concept of artificial intelligence, as far as an artificial device that can be built to think like a human, that goes back until at least ancient Greece, where the Gods (particularly Hephaestus) were said to be able to create animated metal statues that could act and think like people. (Even Homer’s Iliad described them.) But the actual term “artificial intelligence” came out of a workshop at Dartmouth in 1956, and was conceived of by John McCarthy. Science fiction writers were pretty quick to adopt that term into their stories, but it didn’t originate with them. The idea of what AI means does seem to be based on its depiction in science fiction. And to be fair, the workshop that conceived of the term AI had the overly optimistic idea that they’d be able to have true thinking machines that could act just like a human within a generation, and while of course they didn’t even come close to that, they did make a number of advancements that helped improve the computer industry of the day.
I’ll also note that Alan Turing, when he conceived of his Turing Test years earlier, did so in a paper titled, “Computing Machinery and Intelligence”. Because again, AI as a term was not coined yet.
I used “Hide Details” because I’m not quite sure how relevant this is to the discussion, even though the idea was introduced by the OP in opening this discussion.