A battle royal is brewing over copyright and AI
economist.com
economist.com
Somehow we have this notion that we can put an idea (a song, a painting, a text, a meme, you name it) out there and by virtue of copyright people will be obliged to pay for it but will not be able to use it in any way. This is not possible, and advancements in tools such as computers and AI generators make this very obvious. You will NEVER be able to prevent people from mimicking your style, your choices, even your content - you can only prevent them from publishing identical copies and claiming them as their own for a while. I think that's all perfectly fine.
If you don't want your idea to become part of the mountain of content that our civilization is built on: don't publish it. Write it down, store it in your cellar, and have your heirs burn it after you die.
Society already reflects my viewpoint. Culture is based on remixing ideas. Most commonly copyright applies to direct reproduction, although there are obviously degrees. Whether copyright deserves to exist as it currently does is another discussion.
> instead your spouting nonsense just to serve your self interest. Hopefully not though!
What about my position is nonsense, and what do you think my self interest is?
Edit: Look, I'm sorry for rubbing you the wrong way, obviously we both have some strong opinions about AI generators, and maybe tool usage in general. I would like to know more about where you're coming from to have a reaction this visceral. The most glaring point of critique against current AI generators I would understand is the fact that multi-billion dollar startups are printing money using the work of other people for free. Is that it maybe? Would you have a reaction that strong if the models were open source?
There are points of strife of course, the very extended copyright windows we're seeing, the concerns over sampling in music, (particularly a few decades ago when the technology to sample became widely available, though mostly resolved at this point).
I'm not super familiar with all disciplines, but I know within art/drawing, there is a ton of discourse and training to communicate to new artists what appropriate study/remix/use of reference is, and what crosses the line. While there are different viewpoints, I have confidence that for the most part these discussion are based on good faith arguments in ways that serve the artists and those that consume art. Not saying any of this is perfect or couldn't be improved, but the institutional knowledge is significant and dismissing and overriding is probably going to be a lot of work.
On the flip side, there are those that 'justify' whatever they are doing as... well, as whatever argument suits whatever they are doing. This commonly happens with piracy ("I don't want to support big-co XX", or "DRM is evil" or "This is just for personal use"). Any of these would hold a lot more water if anyone who had this opinion actually did any work to to solve whatever problem they were having, while respecting the creators interest (maybe by say, strengthening IP laws so DRM was unnecessary), but that almost never seems to be the case, because turns out the argument isn't actually an argument to modify the concepts of intellectual property to enhance society, it's just an excuse to justify stealing.
This is what I'm seeing constantly with regard to the current round of AI data scraping, people spouting "it's ok because of XX" but then not really having any interest in propagating XX, instead just having an interest in using other people work without paying them to advance their AI product.
Your original post didn't seem to express which camp your interests lie, so I responded as if it could be in either camp.
I take a different tack here, that sometimes would cause me to disregard the creators' interests.
Morally, piracy is preservation. The vast majority of computer software would not be properly preserved if it wasn't for the contemporary pirates. Companies like Nintendo and Apple are not the greatest stewards of software on their platforms, and there are entire classes of works (prototype/development versions of software) that wouldn't be well-preserved in any case.
Plus, respecting the creators' interests in all cases would also allow creators to demand that a work be lost to time, and quite honestly I don't philosophically or morally think that creators should even have the right to do that in the first place, once they have released a work to the public.
Anything less than unabashed piracy risks letting works die. By "die", I don't mean "at the bottom of the totem pole in terms of popularity", I mean "no longer exist in the world at all". Look at the various episodes of classic Doctor Who that have been recovered thanks to fan recordings or private collectors of film.
- Archivism is an important goal
- Often cannot be accomplished in a way that meets the creators interests
- Important enough that sometimes it supersedes the interest of the creators
- Therefore, when all these preconditions hold, piracy is justified to accomplish the goal of archivism.
The other way:
- Morally, piracy is preservation
The former is a line of reasoning I think most people could get behind. I think we could work to outline what cases meet the criteria at each stage, handles cases that other goals might have more precedence (for example the right to privacy might trump the right to archive). This is a lot of work. I would expect multiple lifetimes of energy and activism would be necessary meaningfully move the needle on copyright law in a way that benefited society over all.
The later is a dogmatic statement that steamrolls over any other interests. While there are clearly exceptional cases where the current system fails (that you point out), it's not clear that clinging to this position above all others will result in a net benefit.
It's hard to imagine an expert in Archivism doesn't have some understanding of the moral boundaries of their field. As such, I would expect most experts to take the former approach.
So when the second approach is _exactly what you say_, it's hard to take the "Piracy is preservation" as a meaningful stance to advance Archivism, and instead leads me to interpret it as talking point to justify piracy.
And this isn't to say an individual _couldn't_ start from a dogmatic position, do a ton of work and make the world a better place. Stallman for example has taken a pretty extreme position with free software, done a ton of work, contributed to ton of valuable institutions. All this swimming upstream into much of the rest of societies existing policies. But it seems the exception. And I think in most of these cases the burden of 'demonstrating net societal value' correctly falls on the activist, for example in areas other than free software, Stallmans dogmatic positions haven't been well received and I think appropriately so. (And to be fair, I think most great activists have some of this approach, at least in the storied version of their work, but as Ziwe would say, "How else do you compare yourself to Martin Luther King?")
So yeah, if Archivism is your life's work, good luck. I hope you can find ways to increase the net value of society.
But really I think most people just want free stuff and will say whatever to justify it.
Yes, sometimes piracy requires cracking, and that is not exactly 1:1 preservation. However, that's merely analogous to the difference between the original Doctor Who videotapes that were wiped, and the off-air U-Matic NTSC recordings that were found. While the latter were not at all the same quality as the original, they were nonetheless useful for preservation, as the color signal was able to be used in many cases with the black-and-white telerecording. If one is unable to preserve a work 1:1, it's nonetheless preferable to preserve it in a slightly modified form, than not preserve it at all.
If a program only exists in cracked form, that's a travesty for sure, but it's better than the program not existing at all. In that case, it's also more likely that the original might be found, if there is awareness of the program's existence from the cracked copy.
I fear this will only get worse with advances in cryptography. The ones spending big bucks to develop these protection systems simply do not care about preservation outside of their vaults. They would much rather see the old works go away, so they don't have to compete with them.
I believe retrogaming to be an anomaly that would not exist today without the piracy of the past. The emulation and piracy scenes of the late 90s and early 2000s were seminal in solidifying retrogaming as a subculture in and of itself.
That being said, I agree with most of your statements and concerns. I probably don't weigh them quite as highly compared to creators interests as you do, but I appreciate that you are invested in the issue.
I think it'd be very reasonable to have an institution responsible for archiving you could submit works to (maybe must in some cases?) I know in the us to register a copyright you have to submit the work already, though I would guess it could be the work with drm included. Seems like something in that direction could satisfy archivism without needing to resort to piracy.
the main point is that digital 'property' works essentially like culture, by people copying each other.
whereas non-digital property, i.e. real world property, works differently.
this distinction, in my view, changes the valid view for each two sorts of assets: digital, and non-digital.
> on the theory that that incentive encourages the development of creative works and enriches society more than it would have without those barriers
I don't agree with this view. But I'm thinking that there's another distinction between two types of persons and their applicable incentives. Maybe each kind of person needs different sort of incentive?
So in both cases the issue is to ignore a difference (in the way the asset works and) in the way to incentivize persons.
Well, yes, the thing that didn’t exist until after the law was created did not reach back in time and cause the law to be created. I would think that would not need to be stated, but,...
Well, your country is consuming it, it doesn’t seem very imaginary…
Car is just a faster way to let you move from A to B. But laws treat cars and pedestrians vastly different.
A more extreme(but less related) example: a dog is just a mammal like a pig. But in a lot of countries, dog and cat are "special cases" and trading their meat is illegel.
Laws are extremely inconsistant.
I think that's an entirely different discussion to have. I think AI generators are not faster artists, they're generative tools. Given the amount of prompt engineering and output editing, I'd say significant human input and creativity is still involved. Even in cases where no such work is being done by a human, it's just a generator, not a person. We have been making generated artwork for as long as computers exist, it's just that the quality, artistic range, and accessibility have made a jump now.
> Car is just a faster way to let you move from A to B. But laws treat cars and pedestrians vastly different.
I would say the laws treat printing presses differently from computers, but so far we haven't had any laws (that I know of) trying to protect human typesetters from modern DTP software.
> Laws are extremely inconsistant.
True.
It's even more complicated than that. "derivative mimicry" and "meta-qualities" are both subjective value judgements you are making, and you're implicitly asking us to go along with them as given.
Human output is also derivative. AI output also has meta-qualities. The real difference is that one is made by a person, the other is made by a person using a tool.
> To replace this process with an ML implementation devalues humanity
Maybe there will be a time where machines will - without any human action involved at all - generate content, put it online, autonomously interact with other machine-generated content, and there will be a closed system where humans are either not involved or are just passive consumers. In that case, yes, humans have clearly fallen completely by the wayside. Which, depending on the nature of those interacting autonomous systems, may be the start of a whole new civilization. Or just electrons chasing each other in circuits that are endlessly recombining the ghostly echoes of their long forgotten makers.
However, that day is not today. We have made more capable content generators and generative tools, and amazingly, have democratized them to a degree.
If you can't stand "consuming content" they're useless. So is 95% of modern TV, even if human made, because they're also applying patterns they've been taught.
Once in a while something interesting is made by the content mills - possibly by mistake - and pattern appliers, be they human or automation, have nothing to do with it.
A few humans are capable of going beyond applying patterns, current "AI" isn't at all.
The AI does NOT hold a copyright. That’s ridiculous. The person using the AI does. It’s just a tool like a piano or a paintbrush. A piano doesn’t own the copyright to any music. It can’t. It has no rights. It’s a piano.
By convention, we're using the term artist only for people. Midjourney is not a person, it's a tool. Just like a Gravatar image is not made by the artist Gravatar, it's made by a generator, using rules and some input. Midjourney's rules come from all the stuff it saw when it was trained, and its input comes from humans entering text prompts. You could argue the same applies to human artists, but that doesn't make them equivalent just because they have some things in common.
So, Midjourney is not an artist. But is a person entering a prompt an artist or not? I would argue that there are people skilled and creative at engineering prompts, or further refining the generated outputs. It's not NOT art. Then again, that has always been a matter of interpretation.
> But a machine can only copy. Or you're saying a machine can be an artist.
That's a very Hollywood interpretation. Obviously a machine can be an artist, because humans are also machines (unless you're arguing for the supernatural).
As I said above, the common convention is that an artist should be a person. We have long ago decided that a person can use tools, including generative algorithms, to make what is considered art. However, nobody is obligated to appreciate any given artwork, so I don't see what's being lost if we just let this development play out.
> So - with as few words as possible, who is the artist of that picture?
(HN won't let me reply to your question below directly, so I'll add it here). In your opinion, does every picture HAVE to be made by an artist? If your answer is "yes", then the artist is the human entering the prompt, although the artistic work may have been quite minimal. If your answer is "no", or "it depends on the result", there may be no artist.
Let's say a camera takes a picture of a public place every few minutes and puts it online. Who's the artist?
So - with as few words as possible, who is the artist of that picture?
No, they don't.
Nor, if there is at least one artist, does it need to be exactly one. An image can be created without artistry/authorship or with multiple people’s artistry combined, either by collaboration or otherwise.
So you're saying this picture cant be copyrighted?
does it need to be exactly one
Indeed, the artists are the ones who made the picture used in the training set.
given a training image is good for about 1 byte of information, an image would have..... well over a million likely contributors.
Same as any art an artist comes up with, given all the things they have seen in their life leading up to the image.
Placing it in public domain seems like a easy way to stop it being hard core commercialized.
And honestly, given the crazy extensions on copyright laws, it seems pretty fair to do so.
No, if I was saying that, I’d say “It’s not a sufficiently creative work and cannot be said to have an artist”. What I’m saying is that “you created an image with midjourney” (and the same goes for SD, whether base models or something more custom) does not appear to me (though, to be fair, I'm familiar with the broad concept of the test but not the details of any closely on-point case law – or even loosely analogous specific cases – when it comes to this threshold issue) to be a clearly sufficient basis for asserting either that it must or must not have an artist.
> Indeed, the artists are the ones who made the picture used in the training set.
If there is exactly one picture in the training set and all the model can do is reproduce it then, yes, the artist of that picture is trivially the artist of any of the (identical) pictures generated. But that’s not how actual midjourney works. IF there are many pictures in the training set, and all the model can do is reproduce one of them, but different prompts reproduce different images, then, again, the artist of that image is the artist of the identical result, but, again, Midjourney isn’t Google Images with an “I’m feeling lucky” button, so that’s not the right analysis, either.
You assert that humans and “machines” are equivalent as though it’s some known fact that all rational beings would agree with but I think this is rather far from the truth. I think most people would take issue with your assertion that humans and machines are one and the same.
Human beings show actions, behaviors, emotions which don’t appear to be replicated in any “machine” and we don’t even really know how to begin to make machines that feel emotion.
Art begins with emotion and feeling, the machines we have built show no emotion and as such they are not making art. What we have right now is advanced mimicry, that is all.
In my experience, most artists will gladly list out the artists that they've admired and studied from int their training, and specifically say what pieces or styles influenced any particular work. And then most of the time if you were to communicate that influence to the source artist(s) they would be overjoyed that someone appreciated their work so much that they contributed to the success of the new work.
This is all part of training for and being an artist. These boundaries are set more in (artist) culture than in law, and of course there are those that violate those boundaries, but I think for the most part artists buy in to this shared creative experience, particularly the more invested and successful you get within the field.
> produce a work in the style of $living_artist
it could respond
> With respect to $living_artist, I can't generate work in their style, but if you enjoy their work you can support them at $studio.
The whole "living" thing meaning that at some point this protection expires so you can ask for things in the style of DaVinci, Van Gogh, etc.
Say, for instance, that they instituted a mostly mandatory licensing scheme, so that an individual artist had no choice but to allow use of their art as input when creating generative tools. People using art in this way have to pay a rather high licensing fee, but it is not paid to the artist, but to some sort of central copyright office. Huge copyright holders can also pay an exorbitantly high fee (to the same recipient) to opt out of licensing. Win-win-win; Existing copyright holders keep their existing copyrights, only large-ish actors can create new generative tools, new political positions and institutions are created with lots of money flowing in. Of course, artists then get screwed by being co-opted by generative tools which they can never afford to create themselves, and the general public get robbed both of the opportunity of using and creating new generative tools, and of any less restrictive copyright law.
I don't think that is how it went. Napster was brought down by streaming services.
Just like computer game piracy was ultimate brought down by Steam.
And pirating TV shows and movies was -for a while- stopped by netflix.
I would bet that every time some stupid licensing agreement expires and a TV / movie streaming service has to remove content from their platform -- content people reasonably expected to be there, or worse, content that they were in the middle of or in the habit of enjoying -- a pirate may be born from the frustrated customer.
Keep access flowing, and you'll keep your customers.
Spotify launched in 2006