Lately there has been, quite understandably, a growing backlash against so-called "AI" companies. One of the criticisms that pop up frequently are the violations of copyright law that the "AI" firms have been alleged to commit while getting training data for their LLMs, and while feeding the training data in.
Legally speaking, that's a problem for the courts.
But what about those of us who believe that copyright law, or rather the concept of copyright itself, is heavily heavily flawed and in need of great reform? To what extent are the critics' criticisms valid? I thought it would be nice to write out a very abbreviated argument of how the actions of the aforementioned companies should be taken in our ideal world. Then afterwards, I'll throw in some updates about this blog and just generally about your favourite German state currently inhabiting a human body.
Richard Stallman has an excellent article about why "intellectual property" (sc. copyright + trademark + patents) is not a good way to analyze things. He is right, and his points are very strong especially for trademark being quite different from the rest. Unfortunately, I'm going to conflate patents and copyright in this post, since they do have similarities.
To have a basis for judging the actions, we must first see why patents and copyright exists in the first place. It's quite simple. There is a belief that the "creators" of something should be able to derive benefit from it. What is less simple is in what manner they should benefit.
I wager that the following two rights are the most important, and quite reasonable for "creators" to ask for:
- The right to ask for credit when their work is either reproduced or used as part of a derivative work
- The right to ask for some "reasonable" share of any revenue from reproduced or derivative works
Now, the second one is worded quite carefully. If there is no revenue, there is nothing to give to the "original" creators. And how long should the second right be valid for? Most people would agree 70 years after the death of the creator is pretty absurd. But the exact length would be a subject of much debate. This post is not about that debate, so we'll leave that issue be.
More contentious is allowing the copying and distribution of works without paying anything to the creators if there is no revenue made. If you aren't one of the people who think this is a good idea, this post is not really meant to convince you, because again, this is supposed to be a shortish post, not a 50 page manifesto.
So now, pretending that we have justified that those two rights are what copyright and patents should be about... what? Fine, let me provide some very brief reasoning, while flippantly treating any counter arguments.
Copyright and patents are exist because we care about fairness and economical incentives. Creators, we say, deserve to benefit from their work. It wouldn't be fair for some third party to "steal" the work and profit from it. If we didn't let them profit, why would they ever share their works with the world? Then after some amount of time, it enters the public domain and becomes the property of everybody, so everyone wins!
While this is a perfectly reasonable line of thinking, "IP" law as it stands cannot be explained solely by that basis.
How is it fair to have the right to deny reproductions, translations, distribution, and public performance of a work long after the creator's death[0]? Who does it benefit? Theoretically, the inheritors, who have done nothing[1]. But of course, the reality is that often it is middlemen taking advantage of (or even creating) market inefficiences that capture most of the profit; not the creator or any one of their party.
How is it fair for a work, influenced and defined by countless previous works from the public domain (even if only as an opposing reaction to what came before), to remix the influences, add more or less "original" thought, and demand all the profit? Is the moral thing to do not to pay back our immense debts to society by allowing folks to freely benefit from our works, regardless of their economic means?
How is it fair to hold the whole of humanity hostage with a patent or copyright monopoly? The world we live in right now is built on patent violations by American manufacturers in the 1800s, and more recent IP "theft" by Chinese firms. If both countries were burdened by the full weight of patent law, how much worse would our "standard of living" be right now? How many decades (yes, decades) would that world be lagging compared to us?
Who is being harmed when a patent or copyright is violated, when credit is given, and no revenue is made? Copyright apologists will tell you that this is financial theft, because the potential, theoretical income of the creator is being reduced. This is called "greed". Many things reduce the hypothetical income of the creators. Like when 20 people check out the same library book in a year. Or when you watch a TV show and recommend someone else not watch it. Or...
Ok, you get the point. I will say it: There is nothing morally wrong with gathering petabytes of books and papers and drawings and photos. There is nothing morally wrong with then feeding that into the training process of a LLM.
What is wrong[3] is the lack of credit and profit-sharing. Unfortunately, it seems technically difficult to do either when it comes to LLMs. What a moral LLM company could do is open source the weights, training data and code. Though open sourcing the training data is currently illegal, but the other two are doable. But... how will they make money? Dunno, sell inference? Make custom models? Not our problem, nobody forced them to scrape all that data.
That turned out a little shorter and more comprehensible than I anticipated. A great success. Now to the news:
- My Latin learning is going well. It is a very nice beautiful good language. I am currently writing out parts of the gospel of Matthew. There'll probably be a post about that.
- I've been wanting to write some posts about monks for a while now. I have the concepts of a plan for an overly ambitious multi-part series.
- An addendum on the legally distinct prussianbluehedgehog group's notes on manga translation is almost written up. It'll go in the archives since its not much new, and linked in the original.
- A little bit of work was done on ming-wm. I made malvim (where I am writing this post in) a little bit better. And it turns out id3v1 tags (for mp3 metadata, like song title and artist) are very simple! So I got rid of that dependency from the audio player. Yay! I program much less nowadays. Not really because of AI, ming-wm is just too cool to top, given I am not interested in trying to write a whole operating system... Plus, you know, sometimes I like to do other things.
===
[0]: Moral rights are usually non-transferrable and gone with the death of the creator, so we'll leave it be for now.
[1]: Not to mention the author has plenty of time while living to distribute the profits to their inheritors or save it to pass on if they so chose to.
[2]: Yes, I know fair use exists. It is far too narrow.
[3]: Okay, there's a lot more wrong with LLMs and their effects. But the scope here is only about copyright.
[4]: It sucks that programmers, artists, and all sorts of other professions would get economically outcompeted in many cases, open model or not. It's going to be bad, probably. But again, out of scope. And also, while this sort of thing is awful and comparisons to the industrial revolution putting millions out of work are obnoxious, this seems to be inevitable. Even in professions protected by licensure, the professionals themselves are using "AI". At the end of the day this is the problem with tying your art to money. Realistically, I understand why creatives do it, even when the art form isn't something expensive like filmmaking. But if there is any chance to decouple ourselves from that system and make art without worrying about the economic or clout side of things, and there often is, we should take it! n0thanky0u has a post expressing a more extreme view than mine, regarding something similar, titled "Fuck Artists". I don't agree with it, but it is worth reading.
Lately there has been, quite understandably, a growing backlash against so-called "AI" companies. One of the criticisms that pop up frequently are the violations of copyright law that the "AI" firms have been alleged to commit while getting training data for their LLMs, and while feeding the training data in.
Legally speaking, that's a problem for the courts.
But what about those of us who believe that copyright law, or rather the concept of copyright itself, is heavily heavily flawed and in need of great reform? To what extent are the critics' criticisms valid? I thought it would be nice to write out a *very* abbreviated argument of how the actions of the aforementioned companies should be taken in our ideal world. Then afterwards, I'll throw in some updates about this blog and just generally about your favourite German state currently inhabiting a human body.
Richard Stallman has [an excellent article about why "intellectual property" (sc. copyright + trademark + patents) is not a good way to analyze things](https://www.gnu.org/philosophy/not-ipr.html). He is right, and his points are very strong especially for trademark being quite different from the rest. Unfortunately, I'm going to conflate patents and copyright in this post, since they do have similarities.
To have a basis for judging the actions, we must first see why patents and copyright exists in the first place. It's quite simple. There is a belief that the "creators" of something should be able to derive benefit from it. What is less simple is in what manner they should benefit.
I wager that the following two rights are the most important, and quite reasonable for "creators" to ask for:
1. The right to ask for credit when their work is either reproduced or used as part of a derivative work
2. The right to ask for some "reasonable" share of any revenue from reproduced or derivative works
Now, the second one is worded quite carefully. If there is no revenue, there is nothing to give to the "original" creators. And how long should the second right be valid for? Most people would agree 70 years after the death of the creator is pretty absurd. But the exact length would be a subject of much debate. This post is not about that debate, so we'll leave that issue be.
More contentious is allowing the copying and distribution of works without paying anything to the creators if there is no revenue made. If you aren't one of the people who think this is a good idea, this post is not really meant to convince you, because again, this is supposed to be a shortish post, not a 50 page manifesto.
So now, pretending that we have justified that those two rights are what copyright and patents should be about... what? Fine, let me provide some very brief reasoning, while flippantly treating any counter arguments.
Copyright and patents are exist because we care about **fairness and economical incentives**. Creators, we say, *deserve* to benefit from their work. It wouldn't be fair for some third party to "steal" the work and profit from it. If we didn't let them profit, why would they ever share their works with the world? Then after some amount of time, it enters the public domain and becomes the property of everybody, so everyone wins!
While this is a perfectly reasonable line of thinking, "IP" law as it stands cannot be explained solely by that basis.
How is it fair to have the right to deny reproductions, translations, distribution, and public performance of a work long after the creator's death^\[0\]^? Who does it benefit? Theoretically, the inheritors, who have done nothing^\[1\]^. But of course, the reality is that often it is middlemen taking advantage of (or even creating) market inefficiences that capture most of the profit; not the creator or any one of their party.
How is it fair for a work, influenced and defined by countless previous works from the public domain (even if only as an opposing reaction to what came before), to remix the influences, add more or less "original" thought, and demand all the profit? Is the moral thing to do not to pay back our immense debts to society by allowing folks to freely benefit from our works, regardless of their economic means?
How is it fair to hold the whole of humanity hostage with a patent or copyright monopoly? The world we live in right now is built on patent violations by American manufacturers in the 1800s, and more recent IP "theft" by Chinese firms. If both countries were burdened by the full weight of patent law, how much worse would our "standard of living" be right now? How many decades (yes, *decades*) would that world be lagging compared to us?
Who is being harmed when a patent or copyright is violated, when credit is given, and no revenue is made? Copyright apologists will tell you that this is financial theft, because the potential, theoretical income of the creator is being reduced. **This is called "greed".** Many things reduce the hypothetical income of the creators. Like when 20 people check out the same library book in a year. Or when you watch a TV show and recommend someone else not watch it. Or...
Ok, you get the point. I will say it: **There is nothing morally wrong with gathering petabytes of books and papers and drawings and photos. There is nothing morally wrong with then feeding that into the training process of a LLM.**
What *is* wrong^\[3\]^ is the lack of credit and profit-sharing. Unfortunately, it seems technically difficult to do either when it comes to LLMs. What a moral LLM company could do is open source the weights, training data and code. Though open sourcing the training data is currently illegal, but the other two are doable. But... how will they make money? Dunno, sell inference? Make custom models? Not our problem, nobody forced them to scrape all that data.
That turned out a little shorter and more comprehensible than I anticipated. A great success. Now to the news:
## Various Updates
- My Latin learning is going well. It is a very nice beautiful good language. I am currently writing out parts of the gospel of Matthew. There'll probably be a post about that.
- I've been wanting to write some posts about monks for a while now. I have the concepts of a plan for an overly ambitious multi-part series.
- An addendum on the legally distinct prussianbluehedgehog group's [notes on manga translation](https://www.prussiafan.club/posts/manga-translation-one-rule) is almost written up. It'll go in the archives since its not much new, and linked in the original.
- A little bit of work was done on ming-wm. I made malvim (where I am writing this post in) a little bit better. And it turns out id3v1 tags (for mp3 metadata, like song title and artist) are very simple! So I got rid of that dependency from the audio player. Yay! I program much less nowadays. Not really because of AI, ming-wm is just too cool to top, given I am not interested in trying to write a whole operating system... Plus, you know, sometimes I like to do other things.
===
\[0\]: Moral rights are usually non-transferrable and gone with the death of the creator, so we'll leave it be for now.
\[1\]: Not to mention the author has plenty of time while living to distribute the profits to their inheritors or save it to pass on if they so chose to.
\[2\]: Yes, I know fair use exists. It is far too narrow.
\[3\]: Okay, there's a lot more wrong with LLMs and their effects. But the scope here is only about copyright.
\[4\]: It sucks that programmers, artists, and all sorts of other professions would get economically outcompeted in many cases, open model or not. It's going to be bad, probably. But again, out of scope. And also, while this sort of thing is awful and comparisons to the industrial revolution putting millions out of work are obnoxious, this seems to be inevitable. Even in professions protected by licensure, the professionals themselves are using "AI". At the end of the day this is the problem with tying your art to money. Realistically, I understand why creatives do it, even when the art form isn't something expensive like filmmaking. But if there is any chance to decouple ourselves from that system and make art without worrying about the economic or clout side of things, and there often is, we should take it! n0thanky0u has a [post](https://n0thanky0u.neocities.org/fuckartists) expressing a more extreme view than mine, regarding something similar, titled "Fuck Artists". I don't agree with it, but it is worth reading.