Back in October last year, Walled Culture was one of the first blogs to point out the huge impact that generative AI would have not only on copyright but also on creativity itself. Since then, the world seems to have split into two camps. One believes that generative AI will revolutionise everything, and create some kind of golden age; and the other that thinks the whole thing is a complete sham and/or will destroy civilisation.
The new AI systems certainly have massive problems, not least in the sphere of privacy, as I have written about elsewhere. But the response by the copyright world to generative AI is increasingly extreme, rather as a Walled Culture post back in February warned it might be. The latest manifestation of that tendency is a "Call for Safeguards Around Generative AI in the European AI Act" from "over 40 associations and trade unions that joined the Authors' Rights Initiative". It is a typical anti-technology, anti-progress set of demands from the copyright industry. Its signatories "demand" regulation of generative AI, and they demand it "NOW" (sic).
The document throws in just about every recent criticism of generative AI, some of them undoubtedly quite justified. But those criticisms are largely beside the point, because the letter is really about one thing: copyright, and shielding it from the latest technological advances. [...]
[...] the new document has an entire section devoted to what it calls "The EU's misguided text-and-data mining exemption". Part of it tries to address the argument (made by this blog too) that "use of copyright protected material to train generative AI should be permissible because such training would be equivalent to the (lawful) use of works to get 'inspired'":
The great art and music schools of the world do not train their students on stolen works. Moreover, for humans it takes years of intensive learning, observation of past creations and practicing to master their craft. At the end of such "training", there is a human creative that can make contributions to cultural diversity with individual works that merit copyright protection. This in turn justifies allowing humans to be "inspired" by existing works. In contrast, AI foundation models use computing infrastructure and algorithms to create output that is similar to the training material, in record time. Despite the "training" such systems do not understand what they are doing.
The main argument here is based on the loaded and incorrect term "stolen". Of course, nothing is "stolen" when a digital copy is made for the purpose of machine learning. This is an old rhetorical trick of the copyright industry, and its use here underlines why it should never be accepted.
The fact that it takes humans "years of intensive learning" to master a craft, which is better than computers that can do similar things in "record time", is a curious thing to say, and seems to be arguing that slow is somehow inherently better than fast. Perhaps the signatories of the letter want to get rid of modern agriculture too, because it carries out tasks in record time, and lacks "years of intensive learning" to master the craft of farming. Similarly, the idea that people cannot be inspired by generative AI works because they were produced in a different way is narrow-minded and backward-looking. People can be inspired by anything, however it was produced – look at Andy Warhol's output – so surely the more the merrier? [...]
[...] AI art won't substitute human works except those involving more mundane and unrewarding tasks, and that should surely be welcomed. It's the same process that saw printing presses replace the mind-numbingly slow and boring process of copying texts manually, letter by letter. To be sure, some quill-wielding scribes became redundant, but the arrival of printing presses led to the creation of an entire modern-style publishing industry employing vastly more people. In the same way, generative AI will see the emergence of many new opportunities based on the technology, with millions of new jobs for creative individuals. The only industries that would fear this are ones that refuse to evolve and embrace it.
« IBM Starts Renting Cloudy Bare Metal Linux Almost-Mainframes | China Joins US and Europe in Musing on 3D-printed Moon Bases »
Related Stories
A proposed set of rules by the European Union would, among other things. require makers of generative AI tools such as ChatGPT,to publicize any copyrighted material used by the technology platforms to create content of any kind.
A new draft of European Parliament's legislation, a copy of which was attained by The Wall Street Journal, would allow the original creators of content used by generative AI applications to share in any profits that result.
The European Union's "Artificial Intelligence Act" (AI Act) is the first of its kind by a western set of nations. The proposed legislation relies heavily on existing rules, such as the General Data Protection Regulation (GDPR), the Digital Services Act, and the Digital Markets Act. The AI Act was originally proposed by the European Commission in April 2021.
The bill's provisions also require that the large language models (LLMs) behind generative AI tech, such as the GPT-4, be designed with adequate safeguards against generating content that violates EU laws; that could include child pornography or, in some EU countries, denial of the Holocaust, according to The Washington Post.
[...] But the solution to keeping AI honest isn't easy, according to Avivah Litan, a vice president and distinguished analyst at Gartner Research. It's likely that LLM creators, such as San Fransisco-based OpenAI and others, will need to develop powerful LLMs to check that the ones trained initially have no copyrighted materials. Rules-based systems to filter out copyright materials are likely to be ineffective, Liten said.
(Score: 5, Insightful) by Subsentient on Thursday April 27, @06:34AM (3 children)
AI is inevitable, but there's a right way to do it and a stupid, dangerous way to do it, and it looks like everyone took the stupid way.
"It is no measure of health to be well adjusted to a profoundly sick society." -Jiddu Krishnamurti
(Score: 2, Touché) by shrewdsheep on Thursday April 27, @10:46AM (2 children)
Please enlighten us about the right way to do it.
(Score: 3, Informative) by deimtee on Friday April 28, @12:19AM (1 child)
Elementary, my sheepish friend. If you eliminate all the stupid, dangerous ways to do it, what must be left is the smart, safe way to do it.
No problem is insoluble, but at Ksp = 2.943×10−25 Mercury Sulphide comes close.
(Score: 3, Interesting) by Subsentient on Friday April 28, @05:22AM
This but unironically. Sadly I feel we're too far along to do anything meaningful to regulate it in time.
"It is no measure of health to be well adjusted to a profoundly sick society." -Jiddu Krishnamurti
(Score: 5, Insightful) by ledow on Thursday April 27, @07:31AM (5 children)
FYI - the "copyright industry" includes or represents basically all professional authors, artists, musicians, journalists and just about every other "creative" (including things like website designers, marketing designers, etc.). Also things like open-source programmers, but also commercial programming, etc.
Painting it entirely as some evil corporate robot intent on hoarding all revenue forever and never letting you do anything is really just prejudice (though there are definitely sections of it that are like that).
All the people who argue for no copyright forget that copyright PRECEDED most of modern technology, and was in fact basically invented by classical musicians to protect their work.
Now there's certainly an argument for over-step in the industry and outright abuse in some cases, but I'm not sure how happy you'd be if you spent your life writing your autobiography, it starts selling thousands of copies on Amazon, and then the next thing you know there are a thousand identical copies of it under different names and authors and your revenue goes to zero and Amazon literally say "nothing we can do, you wanted no copyright".
P.S. I don't work "in the copyright industry". I am an open-source programmer (so I do actually give my work away, but under a well-known licence that I expect to be honoured). But I pretty much don't have any art on my walls and I don't listen to music. But I still abide by copyright because I wouldn't want the person who made the things I like to go unrewarded because of situations like the above. Hell, I licenced my *avatar* that I use on certain forums. I checked with the artist, gave them a few quid, and got permission to use it because I liked it.
If we continue to attack copyright like this, things will actually *worsen* because it'll be a copyright free-for-all like China or Russia where there's no point having a good idea or spending lots of time or effort making anything, because you'll never profit personally from it. And it'll be those nasty corporations that you hate that will just steal your work, claim it as their own and use it, as many cases have shown even WITH copyright law (corporations using open-source code without abiding by the OS licence, corporations using people's artwork in their marketing without permission, etc.).
(Score: 2) by turgid on Thursday April 27, @07:43AM
Well said.
I refuse to engage in a battle of wits with an unarmed opponent [wikipedia.org].
(Score: 5, Interesting) by pTamok on Thursday April 27, @08:21AM (2 children)
I kind of agree. Free, libre, open-source software is built on the strength of copyright.
However, my strong opinion is that the duration of copyright is far too long. Economists argue over how to evaluate the 'best' duration, but there are few dissenters from the proposition that existing copyright terms are excessive. My personal favourite would be a term of roughly 15 years.
I would also favour the possibility of copyright extension on payment of an exponentially increasing annual fee, with the bureaucracy required financed by the fees. So payment of 100 bux gives you a year's extension, the following year is 200 bux, then 400 bux, 800 bux, and so on. If you are getting good revenue from something, you can continue to do so for a limited time, but the unprofitable stuff gets into the public domain quickly.
(Score: 4, Interesting) by Zinho on Thursday April 27, @02:56PM
Nor should they be. The past creations that are inspiring our artists-in-training should be freely available in the public domain. The problem in modern society is that the people advocating for stronger copyright protections are wanting to be paid licensing fees by the aspiring artists, and therefore want longer and longer time spans in which they control the copying of said works.
Hear, hear. The infinitely extending copyright terms are a breach of the social contract written into the U.S. constitution, as they steal our past from the general public. Make your money - or don't - in a limited amount of time, then give the artwork to the public domain where it can inspire others to do more.
On a related note, I cringe every time I hear about Disney movies being "put back in the Disney Vault" (i.e. ending sales temporarily to build up demand for a later re-release). I can't think of a clearer way for them to communicate that they are intentionally gatekeeping the public's access to art that should be in the public domain. They aren't even ashamed of it. This is apparently the model that the rest of the "rights holders" aspire to. I can't wait for the day when the public wakes up and pushes back hard enough for legislators to correct the situation, but I'm not holding my breath.
I'm not sure whether hope that the IT sector, which has enough cash on hand to buy the entertainment sector out entirely, will take the matter into their own hands by simply purchasing the companies who complain about their rights being infringed. It would solve the problem for AI development, but unless they immediately donate all those properties to the public domain then it'll be "meet the new boss, same as the old boss." At least the pace of progress in IT is fast enough that they aren't likely to keep pushing for longer copyright terms.
"Space Exploration is not endless circles in low earth orbit." -Buzz Aldrin
(Score: 2) by Freeman on Thursday April 27, @06:06PM
At this point, I would say 50 or 60 years even would be fine. Certainly better than 70yrs+life of author.
Joshua 1:9 "Be strong and of a good courage; be not afraid, neither be thou dismayed: for the Lord thy God is with thee"
(Score: 5, Insightful) by sjames on Thursday April 27, @04:27PM
The problem is, they're not wanting to protect copyright, they want to EXPAND copyright. Copyright says you can't just take their work and run off copies, especially for sale. Fair enough. What they want is a 'right' to say you may not learn from their work or be inspired by their work to create something similar of your own. Currently they are 'willing' to confine that draconian restriction to only apply to people who augment the process with computer simulated neural nets.
Imagine if a craftsman said only I can make a wood carving. You are forbidden to do so, then upon the rejection of his demand says OK, you can do it but only if you don't use any tools, you must carve the wood with your teeth like I do.
(Score: 2, Disagree) by darkfeline on Thursday April 27, @09:35AM (2 children)
That's the entire point of copyright. Copyright and the concept of intellectual property was invented shortly after the invention of the printing press. It was a legal limitation to protect people from losing their livelihoods to technological advancements. That is its essence, to protect against ("protect against" if you prefer) technological progress. The fact that the topic has come up again is entirely expected.
Join the SDF Public Access UNIX System today!
(Score: 3, Informative) by pTamok on Thursday April 27, @10:57AM (1 child)
No it wasn't. Or maybe it depends on what you mean by 'shortly'. Not in England, at least (don't know about China, where printing first was used - woodblock printing for texts on paper originated in China by the 7th century during the Tang dynasty [wikipedia.org]) - it's likely that permission to print was controlled.
The Stationer's Company was set up on 1403 to manage a monopoly on publishing. Once a member had asserted ownership of a text or "copy" by having it approved by the company, no other member was entitled to publish it, that is, no one else had the "right to copy" it. This was not author's copyright, but publisher's. (Paraphrased from Wikipedia: Worshipful Company of Stationers and Newspaper Makers [wikipedia.org]). So copyright started before movable type printing on paper in England. The idea of not being able to freely copy a text and sell it to others - 'intellectual property' - pre-dates printing. It applied to manually produced copies.
(Score: 2) by PiMuNu on Thursday April 27, @11:41AM
See also
https://en.wikipedia.org/wiki/Statute_of_Anne [wikipedia.org]
Note this is 1710, a couple of hundred years after the printing press came to Britain.
(Score: 4, Touché) by RamiK on Thursday April 27, @09:37AM (1 child)
Museums all across Europe are full of looted colonial art and artifacts: https://www.theguardian.com/culture/2019/jun/29/should-museums-return-their-colonial-artefacts [theguardian.com] https://www.euractiv.com/section/languages-culture/news/experts-ask-the-eu-to-step-in-to-return-art-looted-during-colonialism/ [euractiv.com] https://www.swissinfo.ch/eng/culture/looted-colonial-art---there-is-no-limit-to-restitution-/48282022 [swissinfo.ch]
compiling...
(Score: 3, Insightful) by sjames on Thursday April 27, @04:32PM
Beyond that, MANY art students have been seen studying the masters and sketching them while they look. They are not stealing anything, the works were put on public display and they are part of the public. These days many probably do the same looking at a picture of the work online.
The AIs aren't stealing either. The works were available for public viewing online.
(Score: 2) by ilsa on Thursday April 27, @05:27PM (5 children)
Inless I'm misremembering, copyright boils down to ownership of a work. If you create a thing, you (should) have defacto exclusive ownership of said thing. Alternatively, you can grant that ownership to someone else through contract.
So now lets look at AI generated work. A user commands an AI to generate a thing. The AI generates said thing. So now what? AI is not a legally recognised entity, therefore it is incapable of ownership. Similarly, it does not have the legal standing to contractually grant copyright of said work to the requesting user. Works created by AI have no basis for protection in any way.
So where does that leave us? The arguments more or less revolve around the concept that the user using the AI is trying to claim copyright because the AI is "just a tool".
That requires us to define what a "tool" is. The conventional definition is that tools do not have agency. A tool does not have input into the creative process. A tool is a device that facilitates work, through the direct control of the user. A Dremel tool is used by a woodworker to carve or cut. Visual Studio is used by a developer to write code. In each case, it does _exactly_ what the user wants them to do, no more, no less. If the user does the same thing multiple times with the same tool, you will get the same result, just like a mathematical function.
AI does not fall into the above definition because it is not under the control of the user. At best, the user would be classified as a stakeholder who provides some basic requirements and the AI generates it. Beyond that prompt, the user has exactly zero input into the process that creates the final work. If a user provides the exact same prompt to an AI multiple times, the user will _never_ get the same output twice (unless the data set is crap). It is the AI that has agency, NOT the user.
Therefore, the user cannot claim that they simply "used the AI" to generate the work, and that they are the artist and the work is theirs. If a user goes to a woodworker and says, "I need a table of XYZ dimension, with a live edge on one side, blah blah blah", it is the woodworker that did the work. It is the woodworker that had the vision to create that table. If the user then went and said "No, I made that table.", the overwhelming majority of people on the planet would laugh in their face. Replacing the woodworker with AI does not change that equation.
The user didn't create the work, and the AI does not have legal standing to grant the ownership of it's work to the user, ergo, the work in question cannot be owned by anyone at all.
Because copyright evaporates through the Generative AI lens, what you end up with is a vehicle for the artistic equivalent of money laundering.
(Score: 0) by Anonymous Coward on Thursday April 27, @09:23PM (1 child)
Mostly correct --
> If a user goes to a woodworker and says, "I need a table of XYZ dimension, with a live edge on one side, blah blah blah", it is the woodworker that did the work. It is the woodworker that had the vision to create that table. If the user then went and said "No, I made that table.", the overwhelming majority of people on the planet would laugh in their face.
There are (in USA law, not sure elsewhere) specific work for hire contracts. These specify that the buyer ("user" in your example) will own the work product of the maker ("woodworker") and are commonly used when freelancers are brought into projects. The buyer can buy the vision to create the work, as well as the work itself (the "table")...by agreeing on terms in advance.
A search turned up this intro for anyone interested, https://www.legalzoom.com/articles/key-things-to-look-for-in-a-work-for-hire-agreement [legalzoom.com]
(Score: 2) by ilsa on Friday April 28, @06:09PM
What you say is true, but there's one fundamental and critical thing you are glossing over: the word "contract". What is the definition of a contract? A legally binding agreement between two parties.
How does an AI agree to a contract when it isn't sentient, nor have legal rights?
(Score: 2) by sjames on Friday April 28, @03:15AM (2 children)
OTOH, hiring someone to produce a work that you will own the copyright for ( a "work for hire") is commonplace.
I actually have used Stable Diffusion before. You may get lucky, but generally you can NOT just type what you want and the perfect response comes out in a few seconds. It took many iterations to get something reasonably close and then I had to mask out the parts I wanted it to try again and go through a few more iterations.
No doubt AI blurs the lines, but there are tool like aspects to it still.
(Score: 2) by ilsa on Friday April 28, @06:03PM (1 child)
Right, and that's why I'm saying copyright /w AI will be a very steep uphill fight.
The control of copyright ownership is an explicitly contractual thing, and AI by definition does not have legal agency to agree to such any contract. And yet at the same time, it operates like a sub-contractor that does your work for you. So like two kids fighting over a toy and they both let go at the same time, the copyright for a given work falls to the ground and breaks.
It's also why I tried to argue with a hard definition of "tool". One could argue that yes, it's tool-like in that you had to keep trying with repeated prompts until you got what you wanted, but the same is exactly true for your subcontracted woodworker. That's why you go through a long process of back and forth with different sketches, discussions, etc., because it would be too expensive to jump to a final product for every iteration.
As the GenAI offerings improve and you have more control over the generation (eg: controlling whatever seed value, iterative tweaks, etc), then it will more approximate a tool... but at this point I think it's too late. Given the way in which the collective GenAI "industry" released this tech, they shot themselves in the foot so hard with that the gooey meat paste is currently sailing towards Jupiter. I can't even guess how all this is going to shake down, but it's safe to assume lots of lawsuits will be involved.
I'm diverging from the original point of the thread, but IMO this entire copyright nonsense is all just a red herring anyway. It doesn't even cover how much this will enable malicious actors to DDOS any avenue that depends on public interaction. From contests, to scientific paper publications, to gov't requests for public opinion... none of these will be trustworthy anymore because creating massive quantities of bullshit is now mindblowingly easy. We already have a situation where Tesla's lawyers are trying to cast doubts on evidence by claiming prosecuters cannot prove the evidence isn't a deepfake.
One can argue that this technology is the single most dangerous technology ever created. Greater even than nuclear weapons. At least nukes are really hard to make so accessibility is relatively limited (only a handful of entire countries even have them), and there is no disagreement about their danger, making people reasonably second guess their usage. GenAI is not as immediately destructive as a nuke, but it otherwise has no limits to what they can do, no limits to access, no limits to impact, with zero accountability.
Dammit I wrote an essay again. :P
(Score: 2) by sjames on Saturday April 29, @05:50AM
Tools considerably vary in sophistication. For example, what if I produce a work but use various mathematical functions to render the image? If I include random numbers in that? Is the guy who painted by throwing paint into a spinning airplane prop entitled to a copyright? Any of the many artists that use physical processes they have only limited control of (explosives, electrical burning, etc).
I want to distinguish that I in no way believe that the AI ITSELF be eligible to hold copyright as it is not a person, just a sophisticated tool. It lacks will or desire, only doing what it is told.
On a side note, Stable Diffusion offers a fair variety of controls including the ability to enter a seed value, selectively inpaint or re-do a portion of the image, sampling method, de-noising strength, number of steps, etc.
(Score: 1) by SingularityPhoenix on Thursday April 27, @09:48PM
You can still get hand crafted jewelry/pottery/etc. Its just more expensive than whats mass produced by machines. This is just people trying to defend their turf from innovation. A skilled artisan takes years to learn to blow glass, but the injection molder doesn't, and its fine.
They can try to limit it, but like chess engines, they will get more capable, more efficient, and more miniature with time. The cat is out of the bag. There are too many different countries to ban it in all of them, and it will be too easy to get a hold of the software. You wouldn't download a Beethoven would you?
"AI art won't substitute human works except those involving more mundane and unrewarding tasks," I disagree. People will make it churn out whatever it can. What people consider the mundane and unrewarding tasks will shift, and automation changing what is required in these fields.
"The only industries that would fear this are ones that refuse to evolve and embrace it." Oh some industries will have greatly reduced demand, and should be worried. Getting the AI to produce what you want could become an art of its own.
(Score: 2) by jb on Friday April 28, @06:18AM
Even more fundamental than copyright per se is the question of moral rights, most crucially the right of attribution.
Why on earth should the operators of these machine learning systems[1] have the right to plagiarise my (or your or anyone else's) work?
I write free software, which I generally release under an ISC licence, as well as English prose works which I generally release under a CC-BY licence -- in both cases amongst the most permissive licences there are ... but the one thing I will not give up is the right to be acknowledged as the original author of those works.
When someone commits an act of plagiarism it should be the nature of that act itself that determines whether or not it is actionable, NOT the completely irrelevant question of whether the act was committed "by hand" or through the use of some tool (whether ML or otherwise).
[1] ML systems are what they are. Please do not fall into the increasingly common trap of mistakenly calling them AGIs -- real AGI was defined and researched [but never actually implemented -- in fact the general consensus was that it couldn't be] about half a century ago ... and the current "AGIs" certainly do not fall within the term's definition.