It would seem from this article that if a publisher prints a book written by AI the book can not be copyrighted
That means the publisher has no protection.
If that applies also to software, then recent news of AI clones of important apps like Photoshop raises an interesting distinction
the original closed source Photoshop is copyrighted
the AI cloned open source version can not be copyrighted
a human cloned open source version could be copyrighted
Being able to tell whether a work is AI generated might become an important technical activity. Publishers are already struggling trying to distinguish because they understandably do not want to print works they can not copyright.
Software may have an advantage over books here. It is arguably easier to tell if software is AI generated. One technique is to analyse its edit history which should be accessible on github or wherever.
I feel that copyright is far less important for open source software. All it does is preserve the authorâs academic credits.
Over here (The Netherlands) we have a publisher which has opened the spitgots of AI generated books (itâs all they produce). All of them are utter crap, without exception. Recently the publisher started offering pay for humans in order to edit them⌠for such low rates, basically no sane person is willing to apply.
I seriously do hope AI generated stuff gets labeled as copyright violation, without exception or that AI generated stuff cannot be copyrighted.
Like code, itâs easy to figure out whether someone has used AI to create the text; just ask them about it. A person who has used AI to generate his work, almost without exception, does not know his own âworkâ.
I think this human certificate would be a good thing. Have the author prove they did not generate the text or code with AI by providing evidence of their activity. Most people who produce slop, donât care about this.
Also, AI seems to suffer from OCD when itâs generating text or code. Humans are not as consistent as AI.
The big problem is mixing.
If you use an AI spellchecker when writing does that constitute AI content, and does that exclude you from copyrighting it?
If you get AI to generate a book, then modify it by hand, are you able to âownâ it and copyright it.
The law will try to deal with this by setting some stupid cutoff percentage .
The other problem of does AI commit copyright violation in training is easily dealt with. If its knowledge base includes material that it does not hold authors permission documents for, then it has committed copyright violation.
It has done so, and it will get away with it, because the copyright law as it relates to academic work is stupidly restrictive. ⌠there are scientific papers that I can not view online in their original form without paying a fee, for no good reason other than that they were printed and the printer holds the copyright Nothing to do with paying the author⌠authors of scientific work are not supposed to be paid for papers⌠they are paid for doing the work. It is all about printers making money out of making copies.
Other creative areas need to learn from science ⌠pay the artist for work done⌠dont pay for copying rights. The Vatican paid Michaelangelo ⌠but everyone is free to view it.
What is the difference between AI generated and publishers who have a formula for writing books such as mills and boone, or hallmark where all the books are 125 pages, boy meets gorl, boy goes away, girl meets another boy etcâŚ
I have never read any but told that is the formula they use with writers who are just production staff
Very little difference, and publishers could easily replace their team of hack writers with AI. Who would own the resultant books?
We have a neice who is an auditor . She uses AI as if it were a team of unpaid assistants ⌠delegates all the hack reporting to AI and only does herself the initial setting of tasks and final merging of reports.
So, who owns the auditors report?
I think ownership is the key.
I also had a discussion today with an old scientific colleague ⌠what do we get if we ask AI about our own field of work?.
The responses we got were very shallow. ⌠no depth of understanding of the science at all. ⌠because ⌠all AI has to go on is published papers ⌠scientific papers are brief and not ultra-revealing. It might cope better if we waited until someone wrote a textbook on the exact area of research, but that is unlikely in a lot of cases.
So, our conclusion is that AI, as presently trained, is a very shallow view of human knowledge. Most in depth knowledge still resides in humans, not in print or in AI. AI is limited to what printed matter covers.
That is why people still to to University ⌠to be trained by human to human transmission, not to read books or listen to a robot.
To answer that question, and probably others risen here regarding software:
As of today AI doesnât produce anything without a humans request in prior.
And when AI starts to work, eating up a lot of electricity of course, it does it (hopefully) according to the request of the human, this request is called today âpromptâ.
The AI doesnât work for free, tha requestant (the human writing the prompt) has to pay for the burnt tokens.
So my best educated guess (based on all of whatever on I heard on different conferences I was working), that the owner of the AI generated stuff is the human who paid for the tokens.
He/she can do then with it whatever he/she wants. May put in public domain, or may try to sell it, even probably copyright it.
I see a difference in stuff generated with free versions of AI, in such a case I think the result really cannot be copyrighted.
The whole problem is the training data. The question is: does the AI generate something unique, or does it just copy from a billion sources. I think itâs the former, not the latter. In that case it would be infringement on the rights of the author.
Writers already see a remarkably low returns and many of them already cannot live on their work. Whether someone pays an AI company to generate text or not, is not the question here, as the very basis already infringes on the rights of the authors.
I still feel those tech bros committed crimes by committing copyright infringement at the scale they did it. Each and every AI company owner deserves to go to prison, I think. Weâre not talking about one book, or even fifty, weâre talking almost every material human made. Stuff in which people put hard work, theyâre basically claiming as their own through the sale of AI tokens and letting the AI generate content which mixes and matches sources.
Would you be ok with your car getting stolen and somebody putting it for sale as their own?
I dont see why paying for AI responses makes a difference?
Yes if I purchase something, I own it. If I receive a free gift, do I not still own it?
There is a third case ⌠I might own the AI model and run it in my own machine. That is only possible with small models.
True, except for those rogue AI incidents.
There us human input matching every AI output⌠so there is no completdly anonymous AI product.
That sort of answers the ownership question
Neither. It fits a statistical model to text data, then uses the model to predict how a human would respond to various inputs.
Yes they used our text fo fit their model .
It (the model) does not copy , it predicts.
It is the use of texts in fitting their model that may infringe on rights.
I have no,problem with having the whole of humanityâs printed text digitised ⌠but I think it should be available to everyone and be free. We need to find some other way to pay authors. I think the scientific grants system is worth a try.