- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
Why The New York Times might win its copyright lawsuit against OpenAI::The AI community needs to take copyright lawsuits seriously.
Some of the prior cases described in this article, as precedents that could spell trouble for OpenAI, frankly sound like miscarriages of justice. Using copyright to prevent organizations from photocopying articles for internal use? What the heck?
If anything, my take home message is that the reach of copyright law is too long and needs to be taken down a peg.
You are not wrong that monopolies granted by copyright are regularly and unfairly abused.
That being said, AI trainers are getting away with plagiarism right now. More importantly, it’s not just violation of a single copy, it’s potentially the creation of tools that enable mass derivative copies. Authors that create training data need to be compensated.
Authors that create training data need to be compensated.
There should not be a problem with that. The people who work on training datasets are already being paid.
The reason you are getting downvoted is that these lawsuits are not about that. These are about giving money to corporations like the NYT - or Reddit, or Facebook, etc - for the “intellectual property” that they already have lying around. It’s pure grift.
Because the creation of all that is already paid for, that leaves all the more money for lawyers and PR campaigns to extract money for nothing from society.
There should not be a problem with that. The people who work on training datasets are already being paid.
How are the people whose articles and comments are being scraped compensated?
Because the creation of all that is already paid for
“This perfectly good movie has already been made and paid for, that means I can watch it without compensating the studio.”
I do not agree with Reddit selling the comments of their users. Even so that’s a ridiculous statement to make.
How are the people whose articles and comments are being scraped compensated?
By people who work on training datasets I mean, EG, the people on Amazon Mechanical Turk. I am not working on a dataset by writing this comment. I’m putting some things straight and getting exactly the payment I was promised - IE none.
“This perfectly good movie has already been made and paid for, that means I can watch it without compensating the studio.”
Let’s take the NYT as an example. To publish their newspapers, they need to pay reporters, but also editors and assistants. They also need offices, and for those they need to pay maintenance and janitorial staff. To get it out there, they need printers, server admins and such.
In order for this to work, the NYT needs to make back the money that they have paid these people, plus some profit for the owners. This has already been achieved for any issue that’s older than a few days. Before the internet, either an issue sold enough or it didn’t. No one cares about yesterday’s news. I doubt the internet changes that very much. That’s what I mean by “it’s already paid for”.
For a movie, the time horizon is probably a year or so. IDK to be honest. AFAIK, it used to be that if a blockbuster did not make a profit in cinemas, it was over. Maybe the time horizon was longer for direct-to-DVD productions. I guarantee you that no corporations plans ahead more than a few years. Patents last only 20 years and that’s more than enough to finance all the expense for R&D that has created modern tech.
I think it is absolutely ridiculous that corporations can still extract money for something that was made in the 1940ies and even earlier. That does not pay for movies, because it’s not money that was ever calculated with. It only pays for the creation of paywalls. As long as enforcing a copyright pays for that enforcement, it will be done.
In order for this to work, the NYT needs to make back the money that they have paid these people, plus some profit for the owners. This has already been achieved for any issue that’s older than a few days. Before the internet, either an issue sold enough or it didn’t. No one cares about yesterday’s news. I doubt the internet changes that very much. That’s what I mean by “it’s already paid for”.
So the moment a property breaks even, + makes “some profit”, you should no longer need to pay for it? Only when people still “care”, in that case they should pay?
Just because it’s a news article or a comment doesn’t mean it’s fair game all of a sudden.
And movies can make back their budget in the opening week(end) when they’re popular. The timeframe is irrelevant for your argument. At least if we’re talking about anything less than a decade or two old, because…
I think it is absolutely ridiculous that corporations can still extract money for something that was made in the 1940ies and even earlier.
… with this I do agree.
So the moment a property breaks even, + makes “some profit”, you should no l onger need to pay for it? Only when people still “care”, in that case they should pay?
That’s not what I wrote, is it?
The problem with your idea here is that some movies/games/etc never make back the investment. That would mean that they would never run out of copyright if we did it that way. That some movies are duds also means that, on average, the rate of return on such investments is dragged down.
In a functioning market, the average ROI should be the same across the board. If something has a lower return, then people simply don’t invest in it. That’s clear, I hope. This means, that putting a cap on the profit that may be expected will reduce investments.
Obviously, the only returns that matter for this reasoning, are expected returns. Only the expected returns fund movies, etc., and that’s why the timeframe matters.
At this point, ideology (or philosophy) becomes important. One has to ask: What is property about?
There are different philosophical views around this subject, but I am really only concerned with the practical outcome. The political right tends to hold an expansive, absolute view of property rights, to the point of rejecting taxation as illegitimate. The original definition of the political right was as supporters of the monarchy. It makes sense that the right would morph into something that supports all kinds of heritable privilege or right. The anti-capitalistic right seems to have largely disappeared. They often don’t agree that intellectual property is property.
The left tends to hold more nuanced and pragmatic views. Property rights are balanced against other rights; the interests of other people and society at large. The US Constitution takes this view of copyrights and patents. [The United States Congress shall have power] To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.
This latter view is, more or less, one to which I subscribe. Without copyright, there would be only public domain. Copyright integrates creative works into a capitalistic system by turning them into capital. Intellectual property enables people to make a profit with intellectual products. However, there is a clear limit to how much profit one may extract. One may only expect that profit, which actually incentivizes intellectual production. I actually hold the general view that all (commercial) property is only legitimate as long as it works beneficially for society.
So, I do not believe that anyone is entitled to windfall profits. I have published stuff on the web for my own reasons. Other people have found a new use for that by creating AI training datasets. I do not believe I have any moral justification to demand a share of their work.
I hope that clears it up.
Clearly you do not agree with this view. Obviously, you have some more absolutist view of intellectual property. I would appreciate it if you laid out your view of things. You don’t need to answer these question, I’m just putting them here to say what is unclear: How does one create or obtain intellectual property? What can be intellectual property and what are the limits? To what does this property entitle one?
Finally: How come that these threads bring out so much support for right wing views? Looking at other threads, I would have expected left wing views to dominate. Looking at the piracy community around the corner, I would have thought that even among the right, copyright abolitionist views would dominate here. What gives?
That’s exactly what you wrote?
In order for this to work, the NYT needs to make back the money that they have paid these people, plus some profit for the owners. This has already been achieved for any issue that’s older than a few days. Before the internet, either an issue sold enough or it didn’t. No one cares about yesterday’s news. I doubt the internet changes that very much. That’s what I mean by “it’s already paid for”.
Your argument was that the sources that get scraped have already been paid for. I don’t see how it’s any different for newspapers than it is for movies and such. It’s not like news agencies are eternally profitable and never go bankrupt. Nor do I want corporations to profit for free off the comments I wrote, even if I may or may not have signed my soul away in some EULA nobody reads.
AI trainers are getting away with plagiarism right now.
No. They fucking aren’t. 🤦🏼♂️
It looks like someone hasn’t seen the video of Copilot spitting out the Quake inverse sqrt algorithm verbatim.
While it got popularised as “Carmack’s reverse” the algorithm is actually significantly older.
Also you’d have to show that it was literally copy+pasted, including comments and all, to even have a chance at a copyright claim: Algorithms are not subject to copyright, similar to how story structures aren’t. This is like saying “I asked an author to write a book and they plagiarised the hero’s arc!”. And even if it was copied straight-out you’d have an uphill battle to fight, to wit, wikipedia is quoting the thing verbatim.
That said copilot seems to be severely over-fitted in places, and I don’t like the thing one single bit, and the only thing it’s generally good at is writing code faster that shouldn’t have been written in the first place, but inverse sqrt isn’t a good example.
It didn’t just get the gist if the algorithm though, it literally had the same magic number (which isn’t even the most optimal iirc), the same COMMENTS (//what the fuck?), same variable names, etc.
It didn’t produce the algorithm logically, it copied it.
Wikipedia is also adhering to the GPL license of the code. Copilot is not, especially if it’s working on proprietary code or adding an MIT license header to copied GPL code (lol)
I had bing chat spit back at me the question I posted on stack overflow the day before. You know, the example code I provided which didn’t exactly work as I wanted.
It didn’t produce the algorithm logically, it copied it.
The magic number is part of the logic of the thing.
But yes as said copilot is overfitted. Inverse sqrt still isn’t a good example, it’s nearly as bad as Oracle trying to claim to have found copyright infringement in Android’s standard Java library by saying that Math.average or whatnot is identical. There are way better examples of why copilot is fucked up.
The magic number is part of the logic, yes, but that’s not even the best magic number for the job iirc, and nobody remembers how they got it.
I just used this as an example because it’s incredibly clear that it was copied verbatim (again, comments like “what the fuck?” showing up, you can’t tell me it came up with that itself)
A blanket statement is not supported by a single instance. 🤦🏼♂️
“They aren’t getting away with plagiarism”
- “There has been some plagiarism”
“Some plagiarism doesn’t count!”
Your lack of understanding the facts of the situation, much less the definition of plagiarism isn’t a strong argument.
Go on then. If copying a whole function of code verbatim INCLUDING comments like // what the fuck? is not plagiarism in the context of software, what is?
isn’t a strong argument.
Because “no it’s totally not like that” is?
If anything, my take home message is that the reach of copyright law is too long and needs to be taken down a peg.
Exactly! Copyright law is terrible. We need to hold AI companies to the same standard that everyone else is held. Then we might actually get big corporations lobbying to improve copyright law for once. Giving them a free pass right now would be a terrible waste of an opportunity in addition to being an injustice.
I think the photocopying thing models fairly well with user licenses for software. Without commenting on whether that’s right in the grand scheme of things, I can see that as analogous. Most folks accept that they need individual user licenses for software right? I get that photocopying can’t be controlled the same way software can but the case was in the 90s? I mean these things aren’t about whether the provider of the article/software faces increased marginal cost for additional copies/users but that the user/company is getting more use than they paid for. License agreements. Seems like a problem with the terms of licenses and laws rather than how they were judged as following them or not. Their use didn’t seem to be transformative and the for profit nature of their use sort of overruled the “research” fair use.
I also think the mp3.com thing sucks, but again, the way the law is, that’s a reasonable/logical outcome. Same thing that will kill someone offering ebooks to people who show a proof of purchase.
I don’t know the solution to the situation with NYT/open AI. It’s a pretty bad look to be able to spit out an article nearly verbatim. We do need copyright reform, but I think that’s at the feet of the legislators, not judges. I only need to see the recent Alabama IVF court ruling to be reminded of the danger of more… interpretative rulings.
This is the best summary I could come up with:
In its blog post responding to the Times lawsuit, OpenAI wrote that “training AI models using publicly available Internet materials is fair use, as supported by long-standing and widely accepted precedents.”
The most important of these precedents is a 2015 decision that allowed Google to scan millions of copyrighted books to create a search engine.
Stability AI and Anthropic will undoubtedly make similar arguments as they face copyright lawsuits of their own.
But fewer people remember MP3.com, a music startup that tried harder to color inside the lines but still got crushed in the courts.
When a customer wanted to add a CD to their collection, they would put it in their CD-ROM drive just long enough to prove they owned it.
“Defendant purchased tens of thousands of popular CDs in which plaintiffs held the copyrights, and, without authorization, copied their recordings onto its computer servers,” wrote Judge Jed Rakoff in a decision against MP3.com.
The original article contains 644 words, the summary contains 155 words. Saved 76%. I’m a bot and I’m open source!
Ai is really just money laundering but for anything. It takes everything in the models and builds things just like them.
Either we need to abandon copyright or big tech needs to start working on building models that aren’t absolutely full of copywritten content.
Ooh, the first one please. IP law is in a terrible place and the copyright system should be wiped out.