Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Copyright infringement in AI art

technollama.co.uk

Post reply on HN
Print view View original

41–50 of 155 posts

Re: Copyright infringement in AI art

#41

Post by jefftk »

The article gives several examples of AI-regenerated Starry Night, and says:

It looks like it, but it’s not the exact same thing, it’s almost as if the AI is drawing it from memory, which in some way it is, it’s re-constructing what Starry Night looks like.

This seems to ignore that those are pretty clearly derivative works of Starry Night?

Famous UK example which really stretched the definition of a derivative work ("London bus crossing a bridge, black and white") https://en.wikipedia.org/wiki/Temple_Island_Collections_Ltd_...

(Not a lawyer)

Re: Copyright infringement in AI art

#42

Post by sgift »

patengineer wrote: post #37
wesleywt wrote: post #35

If you learn painting by copying other painters then paint a new painting, copyright infringement? Unfortunately many small artists making a living by designing logos will dissappear with this technology.

It depends on what the painter does with the 'new painting'. If it is solely that the painter(considered a student) is learning how to paint, then copyright infringement would not apply as it would be covered under 'fair dealings'. This would apply in the UK, perhaps Europe. I'm not familiar with US law. Quoting the Act - "Fair dealing with a artistic work for the purposes of private study does not infringe any copyr…

Either you or I misread the question. I think it wasn't about selling a copy. It was about using copying to learn how to paint (which is fine, as you've stated) and then produce your own painting which you then sell. The sold painting isn't a copy. So .. no copyright infringement?

Re: Copyright infringement in AI art

#43

Post by tkgally »

OmarAssadi wrote: post #40
tkgally wrote: post #36

Earlier quoted context omitted.

Former professional translator here (Japanese-to-English, 1986 to 2005). I have very mixed feelings about MT. On the one hand, as you and others point out, rates for human translation have been hurt, and the quality of a lot of MT-enabled translation visible to the public is mediocre to poor. On the other hand, MT is enabling communication among people that was not possible or practical before. A small start-up is ab…

Good points! As an example, as a hobby, I am interested in things like anime video compression and such. Two of the most essential sites for me in terms of resources and information have been almost entirely Chinese. And while they appreciate and respect the international users, at the end of the day, they can’t translate each and every single forum post into English or any other language that someone may need. And a…

Thanks for the rant! A bit longer than is normal for HN, but I enjoyed it and learned things from it.

Re: Copyright infringement in AI art

#44

Post by _Algernon_ »

fxtentacle wrote: post #14
gibolt wrote: post #10

Earlier quoted context omitted.

AI isn't locked in a room. No end user can verify the sourced training images, so anything coming close to an existing work should be assumed to have been part of the source material for the derivative.

Oh but you can. I've scanned the LAION training data set (which was used for mini DALLE, Imagen, and Disco Diffusion) and I recognized several images that long to business partners of mine. How does an AI know what Obama looks like? It memorized thousands of images of him, most of which were by professional photographers and, hence, copyrighted.

How does a human know what Obama looks like? They memorized thousands of images[1] of him, most of which were by professional photographers and, hence, copyrighted.

At what point is a neural network sophisticated enough that it can be compared to human learning for copyright purposes?

[1]: For most people at least. Some will have actually seen him in person, and for those this statement would of course not be true.

Re: Copyright infringement in AI art

#45

Post by _trampeltier »

I searched just quick, and for ex. Getty does not allow to use pictures for ML.

Anyway, I think is pretty clear, usually it is not allowed, exept. user upload a picture to a platform and give all rights away because the platform say it in the EULA.

Re: Copyright infringement in AI art

#46

Post by angusturner »

The question that fascinates me about this whole debate is on what basis (if any) one draws the line between a machine learning from copyrighted data vs a human learning from copyrighted data.

And I do think “learning” is the correct metaphor here. There are good reasons to believe that the brain is doing something conceptually similar to generative statistical models. i.e compressing inputs (/percepts) into some kind of abstract representation space. Even if we don’t understand the specific details yet.

As far as I’m concerned, it makes no sense to prevent training on copyright data. If I use a trained model to substantially reproduce parts of a copyrighted work, feels like the issue is largely with me and what I do with the content.

These models are just tools after all.

Re: Copyright infringement in AI art

#47

Post by angusturner »

cgrealy wrote: post #31
stemlord wrote: post #23

Earlier quoted context omitted.

I'm dreaming of an AI that can determine whether a song was AI-generated or is truly one degree from human hands

I highly doubt that’s even theoretically possible, given the vast range of variety and quality of music created by humans.

Feel like people would have said the same with images 6 months ago tbh. (I mean sure, you could get a fake photograph of a person, but the text control that clip has unlocked is just wild…)

Re: Copyright infringement in AI art

#48

Post by anankaie »

angusturner wrote: post #46

The question that fascinates me about this whole debate is on what basis (if any) one draws the line between a machine learning from copyrighted data vs a human learning from copyrighted data. And I do think “learning” is the correct metaphor here. There are good reasons to believe that the brain is doing something conceptually similar to generative statistical models. i.e compressing inputs (/percepts) into some kin…

I suspect that the difference is in knowing versus unknowing use, and the deliberate choice of including learned elements or not on the part of the user.

In the case of the ML model, it can reproduce a part of a training example perfectly, without the user knowing this happened (see the early Copilot analysis wherein it was possibly to get it to quote well known blocks of code verbatim). This is much less likely to happen when a user is consulting her own memories, though obviously never say "never".

Re: Copyright infringement in AI art

#49

Post by trention »

angusturner wrote: post #46

The question that fascinates me about this whole debate is on what basis (if any) one draws the line between a machine learning from copyrighted data vs a human learning from copyrighted data. And I do think “learning” is the correct metaphor here. There are good reasons to believe that the brain is doing something conceptually similar to generative statistical models. i.e compressing inputs (/percepts) into some kin…

On the basis that the "machine" is not a human and there is nothing that necessitates allowing neural networks to do all the things humans are permitted to do.

Re: Copyright infringement in AI art

#50

Post by pornel »

angusturner wrote: post #46

The question that fascinates me about this whole debate is on what basis (if any) one draws the line between a machine learning from copyrighted data vs a human learning from copyrighted data. And I do think “learning” is the correct metaphor here. There are good reasons to believe that the brain is doing something conceptually similar to generative statistical models. i.e compressing inputs (/percepts) into some kin…

I think it makes sense to have a distinction between human and machine learning.

• Human memory has natural limits in capacity, accuracy, speed, etc. Even if you wanted to, you just can't launder copyright all of humanity's works in a blink of an eye. Things done by humans are limited to human scale, but automation can elevate things to an industrial scale (it's legal to fart, it's not legal to have an industrial-scale sewage spill).

• Humans have rights, machines don't. Being able to learn and use knowledge is fundamental for humans, and (pre-Disney) copyright is supposed to be a balance between individual's freedom of creativity and control over their works. Machines don't have needs or rights to learn and express themselves or need to control "their" creative work. Without creator's needs in the picture there's nothing to balance, it's just a business model based on using other people's work.

Post reply on HN