Posted on 06/25/2025 8:45:56 AM PDT by TheDon
In a test case for the artificial intelligence industry, a federal judge has ruled that AI company Anthropic didn’t break the law by training its chatbot Claude on millions of copyrighted books.
(Excerpt) Read more at ocregister.com ...
Most if not all LLMs effectively plagiarize. If we outlaw it, the Chinese will not and will this have a major technological advantage over us.
It’s not outlawing it, it’s about having to ask permission and compensating copyright owners.
It’s how we do things in the West. We don’t steal intellectual property.
If the company pays for a copy of the book, they own it.
Right now, they have the right to use it to train AI—as long as the AI does not use any of the input material in the output.
Copyright protects reproduction of material, not usage for learning.
There are several other cases winding their way through the courts, you can't steal other's intellectual property.
One of the reasons a University degree is expensive is because a student purchases the right to use the course material for their education, but they cannot resell or copy from it directly.
They have stolen our IP for years and they still can't exploit their own petroleum reserves or build a jet engine that doesn't need to be rebuilt after each flight.
I will worry about them when they can do both of these things.
Only if you buy it in physical form and only for certain uses.
If you buy it in electronic form you only buy the right to access it for as long as they allow it and for reasons they approve of.
If Anthropic buys a book and scans it into its training data, this has been ruled fair use by a U.S. federal court. Judge William Alsup determined that training AI models on copyrighted books that were lawfully acquired—such as by purchasing and then scanning physical copies—is “exceedingly transformative” and protected under the fair use doctrine.
“If we outlaw it, the Chinese will not and will this have a major technological advantage over us.”
That is becoming the standard excuse for all the multi billion dollar AI developers. If you don’t let us do whatever we want, whenever we want, the Chinese will win.
Amazing that we are willingly bending over to usher in a system that will be uncontrollable and not necessarily aligned with humanity’s needs or those of the country. If this thing eventually goes rogue, who’ll stop it ???? Nobody because we can’t waste time engineering in failsafe’s for AI because if we do that, somehow the Chinese win. Sheesh.
I don’t disagree with you. I’m actually quite dubious of the value of AI. What AI can do is amazing, but it always has to be tested. Most AI models are trained on the Internet, which we all know is largely untrustworthy. Even models trained on ‘peer reviewed’ scientific journals will spit out garbage, because most peer reviewed journals are full of garbage. AI will only become truly valuable if it can use the scientific method to check its results. I have no idea how that could be done safely or practically. Maybe letting the Chinese have AI is the best strategic option.
They are being sent to trial for their use of pirated book copies. If they had purchased the books they fed to their AI they wouldn’t be in trouble.
But reading a purchased book and digesting the information therein is not a copyright violation, not for a human and not for an AI.
I did not mean to infer anything against you with my original post. It is just that the breakneck speed at which AI is being pushed through is very disturbing to me. It seems that the developers are doing much of this work just because they can and by their own admission this new system will be incredibly disruptive to the existing social order.
There are some (including early proponents of AI) who are warning us about potential pitfalls but it seems those in power are just powering along with no real concern about how this system will displace millions of workers in just a few years or how we can control a system that eventually will be smarter, faster and more responsive than us. A system that can train itself, think for itself, replicate itself and learn and coordinate with other AI and robotic systems. There is no on/off switch.
I find this story especially ironic since it was Anthropic’s own tests that showed it’s own AI system would unethically defy, coerce and mislead the developers when the system itself was informed that it would be replaced.
AI training wasn’t a thing when copyright laws were created. 😉
Hence the need for new laws to ensure copyright owners are compensated.
Not so. They would merely need to get permission and likely pay compensation for the copyrighted material.
That’s the ruling. Which will likely be appealed.
Current law doesn’t cover AI training.
It’s fair to require permission from and compensation to copyright holders. Particularly when the training is used to generate images in the likeness of protected work, i.e. the Disney and Universal case in the UK, and studio Ghibli like images.
It’s easy to see it’s outright theft.
“and not for an AI”
That’s the crux of the matter, if not particularly in this case.
Is it really the same? I think not. New laws need to be created to address this novel use of intellectual property.
Listen to artists and writers on this topic. They’re not happy about it. For them it’s outright theft. It’s not hard to see their point.
Well as an author myself, I find the LLM’s often recommend a quote etc. and generalize the source. Which means it is up to the new author or content creator to properly site the source.
If the book was bought, they have nothing to complain about.
How is AI using it for knowledge different from a human being? Just because AI has better retention and recall?
If I read a lot of copyright material, and develop knowledge from it, should I be prevented from training people by using my knowledge”?
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.