Pearson taking legal action over use of its textbooks to train language models
standard.co.uk
standard.co.uk
So I asked chatgpt about the book, and the exercise, and how to do it, and examples of the 9 cognitive distortions you're supposed to label your responses with. It was very helpful and knew the material well.
It made me wonder this exact point. If you wrote a book on a very rare subject that you were basically the only popular source of detailed info on... And now these models just answer questions for people without having to buy the book.. is that really fair?
Edit: I guess this isn't really that different than the fact I read the book and have described it to many friends and even done the exercise with them.
What if a person bought a copy of the book, read it and learned the material, and then provided advice to people based on the material. I think most people would say that that is fair. So why is it less fair for the AI to do what is effectively the same thing? Is it simply a matter of scale? Or is there something more fundamental at work?
The only point I can see making is that for some reason I doubt openai actually paid the author for the contents of the book, but maybe they did.
Maybe they borrowed ebooks one by one from some digital library and ingested them that way for free.
Much more "scalable" i.e. easy to get raw training dataset. Move fast and break things
Allowing these systems to utilize copyrighted materials without compensating the copyright holder is a legal loophole the size of the sun. Then it becomes a perfectly legal defense to write "an AI" that slurps up copyrighted material and produces results that are "based on" that material.
My issue is that the copyrighted work is fundamentally integrated into the algorithm. Say you wrote a program that would input a copyrighted work, then prints out the contents, but with some words replaced with synonyms. Would the output of that be okay? What about if phrases were replaced? How about if entire sections are added, rearranged, and/or removed? At some point, the results become so mangled as to not be recognizable, but the fact that a copyrighted work serves as an input to the program is still an issue, IMHO.
A major difference between humans and computers is that a human can't (generally) recall hundreds of pages of text and regurgitate it on demand. So a human who has read a book only retains a fraction of the contents, and even then, generally they don't retain all of the information accurately.
My point is not to say that contemporary "AI" systems never emit things that are arguably copyright infringement. It's more just to say that there's nothing about being an AI that makes it necessary that its output be copyright infringing.
But the comparison could just as easily be written, "One is a computer program, literally designed to learn from copyrighted works, the other is a human being that learns from copyrighted works"
If ChatGPT is parroting entire paragraphs of copyrighted work in its answers, then "pilfer" is probably the right word. Or maybe "plagiarize". But if it's training on the information gleaned from the work, and using that training to synthesize the answers, isn't that a lot more like how a learned person applies their knowledge?
I haven't yet made up my mind completely on how IP should interact with AI. I see good arguments for and against. But the answer is far from obvious!
A human can reason, make decisions, act in best interest, judge context of a situation, apply it to new scenarios, judge emotion, and create a new derivative work.
The computer program is literally taking someones work and adding fancy way to search through the other person's work, without adding any value.
We need to stop fantasizing these fancy toys are actually any sort of intelligence.
That strikes me as a rather extraordinary claim. It seems obvious to me that contemporary AI's are "some" sort of intelligence, albeit with open questions around "what kind of intelligence?" and "how intelligent are they?".
Dr Burns seems to have done well, so in his case I'd think he was at least adequately compensated for his life's work. You can't answer "fairness" on copyright in all cases just by that, but this particular case I feel there's little or no harm done.
If people start selling these GPT things because they let you avoid fair recompense, that's another thing we'll have to work out as a society.
Yes.
It's fair because that happens all the time with non-AI intelligences (i.e., people). No author has the right to prevent me from discussing a book with someone. If I were a clinician and incorporated that exercise into my practice, as long as I wasn't copying any content (e.g., worksheets, scoring tables, etc.), then I owe nothing to the author.
It's also a fundamental principle of copyright law, at least in the United States.
https://supreme.justia.com/cases/federal/us/499/340/
https://en.wikipedia.org/wiki/Feist_Publications,_Inc.,_v._R....
a lossy compression function is not "someone"
Facts are not protected by any IP law in the United States.
Ideas may be protected by patent law, if the idea constitutes a novel, non-obvious invention.
The difference is scale. A human can use their learnings from copyrighted work to get a job making $80,000/year. An LLM can use their learnings from copyrighted work to become the biggest, most profitable company ever.
If it helps humanity as a whole, sure. Artificial gate-keeping based on economic factors such as people being able to keep their $80k jobs is not justifiable to me to stop technological progress; one could have said the exact same thing centuries ago with the advent of any number of inventions.
In my viewpoint, something one makes which can be built upon by another freely and without limit is what helps humanity. I don't want anyone necessarily trying to capture rent-seeking "value" out of their work that cannot be extended by others. That is one reason why I abhor copyright, patents, and intellectual property in general.
Money is a way to allocate resources, and technological progress depends on people having incentives to contribute.
Do you think LLMs are a natural phenomenon?
Does it? That stands in stark contrast to open source in general, where people contribute without any monetary incentives.
First of all you added in the word ‘monetary’, to straw-man the position.
Secondly, even if we run with your straw-man, I see little evidence this is true. Most successful projects are either corporately funded or begging for corporate funding. You see article after article here bemoaning the lack of funding for open source.
In today's world with no UBI, it's no wonder people bemoan the lack of funding, since it's an issue of actually living to produce the OSS one wants to make. If we have something akin to UBI, where we don't force people to stall technological progress in order to satisfy having a job, then this issue would disappear.
And to be clear, I'm also not against corporate funding for OSS, so long as it does remain OSS at the end of the day. Linux has corporate funding yet maintains its OSS status.
Which is the truth?
Next question, does this share similarities with academic journals that most folks do consider rent seeking organization?
> In my viewpoint, something one makes which can be built upon by another freely and without limit is what helps humanity.
Now you are saying that LLMs, and maybe even all software, should only be funded by the approval of appointed government committees.
Do you not see the contradiction in your viewpoint?
You aren’t being honest.
Today we’re allowed to fund our own software and use it as we please for private purposes or to license it to others. Today there is no law forcing us to give it away.
This isn’t what you are proposing.
How do you know what I'm proposing? Perhaps we should have a law forcing software to be open source.
> I never said that was okay either. LLMs should be open source and open data.
> That is one reason why I abhor copyright, patents, and intellectual property in general.
If it turns out that closed-source LLMs become the most valuable companies in the world, any thoughts on how to align / reconcile your ideals around the reality of how the technology has developed in my hypothetical example?
Also it seems like you are saying we should just pre-judge cases without using the courts. Basically just end the rule of law. Not sure that’s a good plan.
How would you like to stop this? It doesn’t violate copyright unless they read the book verbatim. I guess ideas could be patented as a utility method by the author and it would prevent someone else from doing that method for 20 years.
Mostly this is just the way things were designed and I think that it’s good in the sense that the purpose of creation is to further human experience, not maximize creator revenue.
I'm sure the author would have no problem with everyone in the world receiving a free summary — once they know they've been fed and cared for.
So what does that mean? They literally asked “Did you read book X during training”?
Sounds like they’re fools. Desperate fools, most likely.