Technology

Well, if you ask e.g. ChatGPT for the lyrics to a song or page after page of a book, and it spits them out 1:1 correct, you could assume that it must have had access to the original.

there are a lot of possible ways to audit an AI for copyrighted works, several of which have been proposed in the comments here, but what this could lead to is laws requiring an accounting log of all material that has been used to train an AI as well as all copyrights and compensation, etc.

Not without some seriously invasive warrants! Ones that will never be granted for an intellectual property case.

Intellectual property is an outdated concept. It used to exist so wealthier outfits couldn't copy your work at scale and muscle you out of an industry you were championing.

It simply does not work the way it was intended. As technology spreads, the barrier for entry into most industries wherein intellectual property is important has been all but demolished.

i.e. 50 years ago: your song that your band performed is great. I have a recording studio and am gonna steal it muahahaha.

Today: "anyone have an audio interface I can borrow so my band can record, mix, master, and release this track?"

Intellectual property ignores the fact that, idk, Issac Newton and Gottfried Wilhelm Leibniz both independently invented calculus at the same time on opposite ends of a disconnected globe. That is to say, intellectual property doesn't exist.

Ever opened a post to make a witty comment to find someone else already made the same witty comment? Yeah. It's like that.

Setting aside the obvious answer of "because capitalism", there are a lot of obstacles towards democratizing this technology. Training of these models is done on clusters of A100 GPU's, which are priced at $10,000USD each. Then there's also the fact that a lot of the progress being made is being done by highly specialized academics, often with the resources of large corporations like Microsoft.

Additionally the curation of datasets is another massive obstacle. We've mostly reached the point of diminishing returns of just throwing all the data at the training of models, it's quickly becoming apparent that the quality of data is far more important than the quantity of the data (see TinyStories as an example). This means a lot of work and research needs to go into qualitative analysis when preparing a dataset. You need a large corpus of input, each of which are above a quality threshold, but then also as a whole they need to represent a wide enough variety of circumstances for you to reach emergence in the domain(s) you're trying to train for.

There is a large and growing body of open source model development, but even that only exists because of Meta "leaking" the original Llama models, and now more recently releasing Llama 2 with a commercial license. Practically overnight an entire ecosystem was born creating higher quality fine-tunes and specialized datasets, but all of that was only possible because Meta invested the resources and made it available to the public.

Actually in hindsight it looks like the answer is still "because capitalism" despite everything I've just said.

It's 100% a new problem. There's established precedent for things costing different amounts depending on their intended use.

For example, buying a consumer copy of song doesn't give you the right to play that song in a stadium or a restaurant.

Training an entire AI to make potentially an infinite number of derived works from your work is 100% worthy of requiring a special agreement. This even goes beyond simple payment to consent; a climate expert might not want their work in an AI which might severely mischatacterize the conclusions, or might want to require that certain queries are regularly checked by a human, etc

When you sell a book, you don’t get to control how that book is used.

This is demonstrably wrong. You cannot buy a book, and then go use it to print your own copies for sale. You cannot use it as a script for a commercial movie. You cannot go publish a sequel to it.

Now please just try to tell me that AI training is specifically covered by fair use and satire case law. Spoiler: you can’t.

This is a novel (pun intended) problem space and deserves to be discussed and decided, like everything else. So yeah, your cavalier dismissal is cavalierly dismissed.

This is a little off, when you quote a book you put the name of the book you’re quoting. When you refer to a book, you, um, refer to the book?

I think the gist of these authors complaints is that a sort of “technology laundered plagiarism” is occurring.

I asked Bing Chat for the 10th paragraph of the first Harry Potter book, and it gave me this:

"He couldn’t know that at this very moment, people meeting in secret all over the country were holding up their glasses and saying in hushed voices: ‘To Harry Potter – the boy who lived!’"

It looks like technically I might be able to obtain the entire book (eventually) by asking Bing the right questions?

However, if the output of an AI would not be considered infringing for a human, then it isn’t infringement.

It's an algorithm that's been trained on numerous pieces of media by a company looking to make money of it. I see no reason to give them a pass on fairly paying for that media.

You can see this if you reverse the comparison, and consider what a human would do to accomplish the task in a professional setting. That's all an algorithm is. An execution of programmed tasks.

If I gave a worker a pirated link to several books and scientific papers in the field, and asked them to synthesize an overview/summary of what they read and publish it, I'd get my ass sued. I have to buy the books and the scientific papers. STEM companies regularly pay for access to papers and codes and standards. Why shouldn't an AI have to do the same?

Because AIs aren't inspired by anything and they don't learn anything

It’s not at all like what humans do. It has no understanding of any concepts whatsoever, it learns nothing. It doesn’t know that it doesn’t know anything even. It’s literally incapable of basic reasoning. It’s essentially taken words and converted them to numbers, and then it examines which string is likely to follow each previous string. When people are writing, they aren’t looking at a huge database of information and determining the most likely word to come next, they’re synthesizing concepts together to create new ones, or building a narrative based on their notes. They understand concepts, they understand definitions. An AI doesn’t, it doesn’t have any conceptual framework, it doesn’t even know what a word is, much less the definition of any of them.

There is nothing silly about that. It's a fundamental question about using content of any kind to train artificial intelligence that affects way more than just writers.

So, in your "only hope for a happy ending" scenario, how do the artists get paid? Or will we no longer need them after AI runs everything ;)

Isn’t learning the basic act of reading text?

not even close. that's not how AI training models work, either.

if your position is that only humans can learn and adapt text

nope-- their demands are right at the top of the article and in the summary for this post:

Thousands of authors demand payment from AI companies for use of copyrighted works::Thousands of published authors are requesting payment from tech companies for the use of their copyrighted works in training artificial intelligence tools

that broadly rules out any AI ever

only if the companies training AI refuse to pay

There's a difference between a sapient creature drawing inspiration and a glorified autocomplete using copyrighted text to produce sentences which are only cogent due to substantial reliance upon those copyrighted texts.

All AI creations are derivative and subject to copyright law.