AI is stealing writers’ words and jobs…

Status
Not open for further replies.
@Snarf Zagyg. Found it, looks like it was just melodies

I don't click through videos, but it appears that this was more of a "thought experiment" than a real legal attempt.

Here is a quick (contemporaneous) legal analysis.

Given what we know of machine-generated works since that point, I would say, quite strongly at this point, that this didn't do anything.
 

log in or register to remove this ad



Okay? I was mentioned in this thread, and responded. I didn't read the prior 380 posts. But sure, you are welcome to the credit!

I will endeavor, in the future, to charge people for writing so I don't get summoned again like some kind of ersatz Beetlejuice.
Oh, I didn't mean it like that. Just that we cited the same case.
 


Forgive me if I don't put much stock in the ruling of AI and copyrights from a country known for stealing IP and selling knock-offs. I put more weight into a US court, which has ruled that you can't copyright AI art.

In fact, even AI companies state you can't copyright anything you produce using their programs.
I know, hence the part i quoted, which is something I've pointed out myself and then been told that doesn't make it art.


I'm aware of the differences but a thought did pop into my head about companies and actual folks who make the publicly available models not paying for content that they use in making the models and the hay day of Napster/lime wire and sites like pirate bay, I feel that and i'm pretty sure there's actual hypocrisy overlap there.
 

You are stunningly incorrect. I most distinctly have NOT said that it was nothing but a repackaged image search.
I didn't mean a literal repackaged image search, I was referring to the false implication, which people - not necessarily you but definitely some people - seem to be making, that images output by AI will be materially interchangable with existing training images more often than not. Not as a fluke, not as a reult of a prompt deliberately describing a training image in detail, but as a rule. That the outputs are as a rule materially the same as the training inputs. There definitely seems to be at least an implication of that.

And if that's not wgat people are saying and they're not interchangable with the training images how could there possibly be any IP infringement?
 


And if that's not wgat people are saying and they're not interchangable with the training images how could there possibly be any IP infringement?

Well, there's several arguments here:

One, made by Sarah Silverman, and others, is that if the AI can accurately summarize or largely duplicate an original work, that is infringement. Basically, if you can give the generative AI a request to duplicate an existing work, the original persists in the system in sufficient form to be considered infringement. This argument isn't doing so hot in the courts at the moment, but some analysis I have read suggests that is because the topic is new, and that as more issues arise, this argument may gain more legs.

Another is a simple, but powerful, technical point that already has legal precedent - in order to use a work as training data, the original work must be copied as digital data in the AI training system! Done without permission, that, itself, is a violation of copyright, which prohibits duplication of covered works. While there is a long tradition of allowing individual consumers to create backups or copies for personal use, doing so for other purposes is another matter.

Remember Napster? This is what ultimately took down Napster. You can't make a digital copy of someone else's work and do whatever you want with it, just because you feel like it. Period.

Early Generative AI work leaned heavily on their being research, sliding in under the educational arm of Fair Use. But that argument ceases to hold as you open distribution of the result to the general public and commercial concerns.

You will note that the training sets used often avoid bodies of work controlled by major business concerns. Nobody has gone and scanned a bunch of Disney animated films and used them as training data, because Disney has the wherewithal to defend itself. Instead, training data is often lifted from the internet, and smaller artists who cannot mount a significant legal defense individually.

This strongly suggests that the folks getting this data know it is legally dicey.
 


Status
Not open for further replies.
Remove ads

Top