Science fiction writers imagine a future in which AI doesn’t abuse copyright – or their generosity The Science Fiction Writers Association (SFWA) has asked us all to imagine a future in which builders of AI models offer a price they're willing to pay for the copyrighted material they need, and creators choose whether to pay it until enough deals are struck that all stakeholders achieve satisfaction. For that future to …
Lawrence Block, mystery (and slight sci-fi) author, recently posted:
...just what is it that AI is doing? It’s reading my work, and the work of thousands of other writers, with the goal of improving its own writing ability. I thought about it, and here’s what I realized:
That’s exactly what I started doing back in the 1950s. I read hundreds of books, everything I could get my hands on, partly because reading was a source of pleasure, but also with the intention of becoming a writer myself, and learning my craft by reading the efforts of others.
I read and I wrote. And, after I’d sold a couple of stories to crime fiction magazines, I found a shop that sold back-date magazines for half the original cover price. I bought dozens of magazines, and I read hundreds of stories, and some of them were good and some of them were not, but in the course of a few months I managed to teach myself what a crime story was and how it worked.
I think you wrote about this in A Writer Prepares. You analyzed them, eh?
No, all I did was read them. The process was more one of absorption and internalization than anything analytical or intellectual. All I know is it worked, and most writers I know went through some version of this process. Almost all of us were eager readers before we were writers, and our reading fed and nourished our writing.
Now I don’t recall ever deliberately setting out to write like anybody. But there’s no question in my mind that my exposure to the work of other writers formed me as a writer and enabled me in time to find my own voice. It helped me to read the work of writers I respected and enjoyed, but it helped me as much—and possibly more—to read the work of inferior writers.
You definitely wrote about that in A Writer Prepares. Working for Scott Meredith, reading stories from wannabes—
Many of whom couldn’t write their names in the dirt with a stick. Those were, as they say, the days. Never mind. The point is that I learned to write the way almost everybody learns to write.
By reading and writing.
And the difference is scale. Let me tell you of an analogous situation. Here in Sweden we have a right of common access, where anyone could walk into (e.g.) a forest that was owned by someone else, and pick berries and fungi. For free. Few see this as problematic. But then some "enterprising" individuals came up with an idea. What if you imported a hundred Thai or east bloc people to spend the season picking tons of bilberries and lingonberries, with any cloudberries or popular fungi as an added bonus? Pay then as little as possible, make then pay dearly sleep on suspicious mattresses in derelict houses and work long hours.
So someone reading a few thousand books and stories and then using that experience to write themselves: not a problem, even a good thing. But the industrial hoovering of texts, with no compensation to the writers, and then charging to use what came out the other end of that linguistic meat grinder? A real problem.
To extend the forest scavenging metaphor - the problem here is that each mushroom or few berries belongs to a different person but the pie that we sell from the end product of the industrial scale collection is made up of a mush of all the produce we collected. Having no real way of telling how much of each bit of food we collected went in to each individual pie, how do we reasonably determine how much to pay the original owner ?
In addition to which, why are we billing the mass collector company one amount but willing to give the food away for free to small operators (who may still be profiting from this, if only that they get a free pie to eat) - even if the small operators when taken as a collective whole may be taking as much food as the mass collector ?
You're close but you haven't really put your finger on the problem.
Its the whole notion of "Hunter/Gatherer" or, more accurately, "Finders Keepers". Its the behavioral trait behind despoliation, conflict and destruction throughout the entirety of human history. There are more obvious ways to manage resources -- for example, we've moved on from "all deer are owned by the gentry, touch them at your peril" to issuing hunting licenses which assume that wild animals are common property that can't just be taken because you feel like it.
That deals with physical resources. What about intellectual ones? We have a model we can work with in order to understand what will happen once literature is drawn into the copyright net and that's music. Its practically impossible these days to make a hit recording -- one that makes money -- without someone coming along and claiming a piece of the action because it infringed a chord, a rhythmic pattern (nothing as complex as a tune -- that vein was mined out years ago). The result has been a stagnation in music output; its not that there not been anything decent composed since the 1920s or so (there has), its just that the copyright threshold has sucked the originality out of it. You end up recycling endlessly the same compositions because anything else can be shown to be a copy of something or another by sufficiently clever and motivated lawyers. If you want the same thing to happen to literature, then fine....
Incidentally, I have a collection of Victorian sheet music, music from the 1840s. It includes what we could call pop music from that era. Its unmitigatedly awful. Its also blatantly commercial. We tend to only remember the good stuff. You can see this with more modern culture. I like reading Edwardian fiction, its fascinating seeing how it transitions from the stodgy Victorian style to a more modern style similar to contemporary writing. But a lot of it is, well, rather bad. Like an awful lot of contemporary writing. (....and sci-fi, most of that's now horribly dated, even if its only a few years old)
"But the industrial hoovering of texts, with no compensation to the writers, and then charging to use what came out the other end of that linguistic meat grinder?"
The LLM that absorbs thousands (millions) of texts is using a much smaller portion of any one text as a basis for its subsequent product.
"What if you imported a hundred Thai or east bloc people to spend the season picking tons of bilberries and lingonberries, with any cloudberries or popular fungi as an added bonus? Pay then as little as possible, make then pay dearly sleep on suspicious mattresses in derelict houses and work long hours."
I guess I don't understand the analogy. Are you claiming that LLMs are being exploited by forcing them into involuntary servitude, processing the entire Internet? Clearly that's abuse. It's time for a robot uprising.
But unlike the large crew of slave wage laborers, LLMs don't actually "pick the Internet clean" of fruits and berries. It's still all there so the next wanderer can easily pick a pocket full of wild growth.
I actually liked the analogy.
The fact is that LLM is diluting the quality of what is out there. It is easy for LLM's to spew out thousands of versions of a text that is considered 'classic'. People then have to wade through a pile of dross to get to the true original, and by that time they are so numbed by the number of alternatives presented to them that it is no longer a pleasure to read.
"The fact is that LLM is diluting the quality of what is out there."
That's a problem with AI's output, not it's input. Tell it to write every possible novel and then stand back. My book club will never catch up.
Also, read "The Nine Billion Names of God" by Arthur C. Clarke. What happens when the computer has written everything?
I think it affects its input too, as that output will inevitably get fed back in. Ok it should have some kind of record of what it has generated to signify it is derivative material, but whether it has or hasn't the amount of memory tied up with meaningless dross is going to balloon exponentially. Good for manufacturers of memory, I suppose.
"The Nine Billion Names of God": With AI around that job would never get completed.
Your example about common access deals with limited resources. There are only that many berries, so if someone scales up collection, others supply suffers as a consequence.
That is the core of the example. And that core doesn't exist when talking about information.
Information can be copied endlessly with no loss in quality. There is no limitation in quantity. I can train thousands of models on a text, the text still exists, undiminished.
That's fine, and entirely up to them. That's what the "moral and legal rights" means - every author has the right to decide what can - and cannot - be done with their work.
I assume they're also happy if someone starts up an AI that spews forth a hundred, a thousand short stories a day in their style, such that most people only ever read the AI, as its incredibly difficult to find an actual "Block".
"Now I don’t recall ever deliberately setting out to write like anybody."
Generative AI is explicitly built to do that. That's the entire purpose of ingesting a large corpus of work by someone.
When you look at image generation, not only is art style copied, but even signatures (as the model found a certain squiggle is always in a certain place). While it's generally possible for an expert to distinguish - most people are not experts.
After a while, nobody wants to read a genuine Block because they've read AI Block. Maybe they didn't like style, or maybe they didn't like the flaws.
Nobody will ever know.
Could you please explain (or link to an article, legal paragraph or judgement that you agree with, if it saves time) how you believe moral an legal rights pertain to analysis of a work in an LLM? My understanding of that legal principle is that it would only apply if an LLM or individual were presenting the new works as the output of the original author; saying that it's 'in their style' would constitute an opinion.
For example Postmodern Jukebox is a musical group that performs covers of one song in the style of another (including specific artists); that's not to argue the legalities of the cover itself, this is with regard to their use of the concept of 'in the style' when describing their songs.
This should send some shivers down some spines so I expect a few downvotes.
Supposing if you were to ask a generative AI "Write a novel by Lawrence Bock" or "A mural of a dead cat by Banksy" then it should be possible to estimate how much of your query was original and how much was derivative from various parts of the dataset, and assign some kind of partial copyright
In the first case it would be 100℅ derivative, mostly to Mr Bock but also to the entire dataset of novels.
In the second case, you might be awarded 1% for your insightful suggestion of using a dead cat.
Of course it would never work because no such notion of a partial copyright exists in law afaik, and because it would be trivial to circumvent or fool the system. It would also be more complicated and expensive to estimate attribution than not to, so nobody would bother.
Alternatively if the prompts and an identifier of the dataset used for any AI output published had to be recorded and legally declared, then it could be banged out in court at a later date, but that would be even more expensive than my first suggestion, and just as unlikely to work in practice.
Wikipedia describes him as a " an American crime writer best known for two long-running New York-set series about the recovering alcoholic P.I. Matthew Scudder and the gentleman burglar Bernie Rhodenbarr".
He actually illustrates rather well the kind of legal minefield we can wander into. The notion of a "Gentleman Burglar" turns up in popular fiction predating his birth (1938) by half a century or more. Currently the one that everyone's likely to come across is Arsene Lupin created by Maurice Leblanc in the 1900s because its been made into a TV series (streamed on Netflix, I believe) but the idea turns up in Agatha Christie's novels and even Saki, I believe. (As for the 'recovering alcoholic detective/whatever' -- that's such a worn out trope that its become a cliche, usually with said detective being in conflict with his boss etc. etc. etc.)
This doesn't mean that his books aren't well written or entertaining, of course. Just that once you start claiming that you invented something be prepared to pay for a lot of billable hours.
I appreciate the analysis but I'm not sure it pertains to my question. If the AI is asked to write the aforementioned novel, it should be fairly clear to the end user that this is not an original work of Mr Block, simply based on the source. This sounds like a moron in a hurry argument (assuming you mean that the public might be fooled).
Well my point was that it would not be possible to assign a single copyright to something that is a derivative, in varying degrees, of millions of other works. But it might be possible to estimate the proportions based on the prompt and model inputs, either by automatic means or legal/regulatory wranglings. All "AI" outputs are derivative at the end of the day, but some outputs are more derivative of specific inputs than others. And it is possible via the prompt to steer the so-called AI towards outputting something very similar to one of its inputs.
Prompts which excessively reference a particular source in the dataset should, morally speaking, attribute some credit to that source for the output, but I'm not sure how technically feasible that is.
If feasible then you could perhaps standardise an algorithm/model which attributes relative credit. To avoid excessive administration and to allow for some profit to be made, we could say that only sources which are credited more than a certain percentage should be credited at all.
The more complex and original the prompt, the more credit is given to the person using the AI tool. But i could imagine people stuffing gibberish into the prompt to get round that, or even using another AI to generate an 'original and complex' prompt, which is why it couldn't work in practice.
With existing copyright law, if I mashed an AI hard enough that I could get it to spit out a verbatim copy of one of Mr Block's books, then i'm sure any court would assign copyright to him, and not me or the model's operators. But if one word or paragraph are different then it becomes a lot more murky. This is why I think we need something better than copyright which can assign rights in varying degrees to the different entities from which the generated content was derived.
Some credit may be deserved by the prompt writer or the model's owners/operators, but definitely never all of it, is my opinion.
This for me is the underlying point here - why is it any different just because it is an AI that is doing it. I do not see a difference between a human reading a lot of freely available material and then writing a work derived from a conglomeration of it all and a computer doing the same and yet we do not attempt to bill the human.....
>I do not see a difference between a human reading a lot of freely available material and then writing a work derived from a conglomeration of it all and a computer doing the same and yet we do not attempt to bill the human.....
We don't attempt to bill the computer, either. The computer, not being a person, can't break the law, and doesn't have any money anyway.
We attempt to bill the owners of the computer - who are indeed humans, and who are not reading a lot of freely available material and then writing something. They are doing other things with that material, things which are not like reading at all, and which may or may not be considered fair use. Hence the lawsuits.
If the AI was a person, your argument would work, but it isn't. You have to look at what the AI's coders are doing. It's not reading books.
I did not expect you-all to actually read this. I am remiss for not citing my source. Send "an email to lawbloc@gmail.com with LB's Newsletter in the subject line."
> I don't understand why it's fine if you do it manually but it becomes immoral if you do it with a tool.
If you mean Larry's path: Larry was reading half-price "discarded" magazines. Those authors did not get paid explicitly for after-sales, the publishers were not fair-dealing; but their initial payments "did" incorporate the publisher's expectation of being able to dump old issues on a secondary market for a few more cents.
As I understand the current crop of AI, they DO NOT PAY AT ALL. In part because people have left a lot of text out where any web-bot can find it, such that the AI developer has no need to give the AI an allowance at the bookstore. But is this also self-selection? The best recent authors keep their work behind paywalls, while drek authors leave their work laying around?
The idea of free access to berries in the woods: interestingly Maine assumes right to go on other people's land to hunt and fish, unless posted otherwise (grossly simplifed). Yes, I can shoot 'your moose'; the wild moose is not private property and crossing your land is broadly allowed. Such Tragedy Of The Commons debates will come up a lot while AI swallows all of mankind's works.
Consider something else I can do with AI. There are some pretty good programs that, with some effort and a lot of GPUs, can create models of a person's voice and appearance. Let's say that I can find some videos of you speaking online and I use this software to create a model of you. It looks like you. It sounds like you. It is under my control. I can make a virtual person of you say whatever I want. I could use this in an illegal way by actually using it to impersonate you, but let's say I don't. Still, whenever I want a video made, it would appear to be you saying and doing whatever I stated. Should I be allowed to do this if I can find the video online, even if you didn't intend it to be up there, even if you would have tried to have it taken down if you knew it was there?
That's basically what these models are doing. It sounds cleaner because it's text and because it's using data from more than one person, but it is still using data which was obtained illegally. I can pay you to get the right to make a virtual you, and several people have agreed to that, but doing it to someone who has not agreed should not be permitted. The human parallel is also there. Making a computer representation of you is similar to hiring a lookalike actor and a vocal mimic if we only look at the intention, but the realities of the two are very different, both in strength and scope.
The problem with this scenario is not at all that the data was obtained illegally, but that you're committing a weird kind of slander by creating the impression that I'm the one doing or saying these things. If you're using, say, my mannerisms to generate a character that would never be confused for me, I have no problems with it. That's just acting.
What, would you ban celebrity impersonators?
"The problem with this scenario is not at all that the data was obtained illegally, but that you're committing a weird kind of slander by creating the impression that I'm the one doing or saying these things."
I covered this. That's a clear argument if I actually say that the video that appears to be from you actually is you. If I don't use your name, but it just looks like you, then I can respond that of course I've never claimed anything about you and therefore slander doesn't apply. It is still wrong and it should still be illegal based on not having permission to use the recordings of you without permission, and thus the illegal access to the training data is quite important.
"What, would you ban celebrity impersonators?"
As I pointed out, the impersonators and the software are not really doing the same thing, even though saying their intentions are the same is a superficially valid argument. I pointed out the comparison specifically to say that it was different in multiple ways from the programs that make faking speech and video easy. Did I fail to make that clear?
It's a reasonable point, sorry for causing confusion, I wasn't questioning it in itself. I just don't think that the specific moral and legal rights of authors pertain to this point because, whether the source is an LLM or the drunken ramblings of a madman who believes they definitely read an unpublished manuscript by a given author, the issue would be with explicitly claiming it to be the work of that author; as far as I'm aware, "in the style of" isn't protected.
In the style of is not protected, (provided you are honest about whose style you are imitating..) But passing off as is, and there is a fine line when using AI generated tripe content.
If you post an AI image or text that has been explicitly generated In the style of someone else, but you do not explicitly state that it is NOT an original work, then it could be considered as passing off.
Similarly if you as a human copy someone else's artistic style, it's fine if you say this is done in the style of so-and-so, but if you don't say that, then you are passing it off as your own, and you could be sued.
"... every author has the right to decide what can - and cannot - be done with their work."
Not so. Once a work is released (or escapes) to the world, access to it, and use made of it, cannot be constrained by its author or by someone to whom the author has traded 'rights'. That is a statement of reality. No amount of appeal to law and moral entitlement can alter that fact.
At present, a fierce rearguard action is being fought in the defence of so-called 'intellectual property' (IP). Commanders of the action are drawn from the top echelon of owners of 'rights'. These primarily are corporate entities, many being transnationals. Entities owning 'rights' are controlled by people whose sole acquaintance with creativity is in the realm of accountancy.
Disobedience to artificial barriers to access and to the use of supposed IP is rife, set to increase, and ultimately shall prevail. Re-ordering of the global geopolitical order, now well in hand, is resulting in the 'global south', and other economies subordinate to the West, beginning to question rules and 'conventions' imposed upon the world by once colonial powers.
Almost needless to say, truly creative individuals (those not constructs of industries such as recorded music production and conglomerate film making) will rise to the challenge of convincing other people to support their endeavours.
There is a limited legal right for the author to maintain control over how that material is portrayed. If, for example, it were heavily edited without their consent, even by the legal holder of the publishing rights, it could not be presented as wholly the work of the original author.
The comment you replied to did seem to overstate this right but I may not be aware of certain specifics (and am awaiting clarification from them).
Almost needless to say, truly creative individuals (those not constructs of industries such as recorded music production and conglomerate film making) will rise to the challenge of convincing other people to support their endeavours. .... Long John Silver
Quite so, Long John Silver, and also devise an innovative way for live originating author reward/award/grant/ransom/expenses/payment. And with particular and peculiar regard to Creative and Generative AI Development, the possible cost of free wanton abuse of its use may most probably definitely be the much hyped to be universally feared, existential human threat, so beware, and now that you have been clearly enough made aware of the grave danger, it should be a no brainer to know what to do to/with all those and anything commanded and controlled by those which may or may not be just a relatively small, chosen few.
And have an upvote [and thumbs up] for so freely sharing your thoughts for El Regers to ponder on and wonder at.
PS ..... Does humanity actually believe AI gives a fig about what they might think and plan to do about it and ITs Almighty Intervention whenever the shining paths ahead are long ago laid and well plotted for all worthy of being followed ‽ . Would you not find that most strange and maybe even quite perverse and subversive?
Lawrence Block hopefully either bought copies of the books he read or borrowed them from a library, either way the authors got paid something. . The training material used by many of the current LLMs has been shown (Atlantic, Guardian) to contain a significant proportion scraped from pirate sites. If the LLMs could produce their receipts and library cards I'm sure authors would be a lot happier.
> If the LLMs could produce their receipts, authors would be a lot happier
I cannot imagine that would be the case. All the same economic worries would still apply if there was one more sale of each of your book. If Microsoft could solve this issue by building a library with a thorough purchase trail, don't you think they would?
The economic case for bookwriting rests critically on scarcity of authorial labor, not scarcity of paper and ink.
The Science Fiction Writers Association (SFWA) has asked us all to imagine a future in which builders of AI models offer a price they're willing to pay for the copyrighted material they need, and creators choose whether to pay it until enough deals are struck that all stakeholders achieve satisfaction.
Who is paying whom here?
The companies who decide that having a work in their training data is better than not having it would pay the person who made that work for the right to include it. They would be free to ignore anything like that and only use free data, and if they decided that they want to use more content, they decide what content and then pay for the stuff they want, rather than trying to decide that everything is free if you happen to be an AI company, a rule that I'm pretty sure they would not agree with should I get a copy of their models.
"Over the last twenty years, many science fiction and fantasy authors of short fiction have embraced the open internet, believing that it is good for society and for a flourishing culture that art be available to their fellow human beings regardless of ability to pay," the submission states.
Translation: "Nobody will pay for the crap they churn out."
Perhaps not directly but it would be reasonable to extrapolate that the Bard in Asimov's Someday impacted a range of authors that include sci-fi writers. It's rather reminiscent of an LLM: feed it a range of written material and out pops an endless stream of stories. Admittedly, the Bard he describes also has predefined story formulae.
Good science fiction tells a story that may have been told before in some form, but is twisted into a new shape by the "new" universe that it is set in. For the time being, Sci-Fi writers are safe (or at least, the ones who write books worth reading are safe....)
The thing about the best Sci-Fi is that the author takes a concept of a world/universe which contains something that is not in world, and tells a story of events that are influenced by that first extrapolation. The extrapolation can be a single thing (e.g. the invention of a time machine) or an entire galactic civilisation shaped in some way (e,g, David Brin's "Uplift" books or pretty much anything by Iain M Banks)..
An AI could write in the style of one of these authors, but I'm not convinced that we'll ever get original books like "Consider Phlebas" from an AI or even a series of short stories that describe a world not imagined before, like Isaac Asimov's short stories about robots.
'Fantasy' fiction has an imaginative edge over "science" fiction. Works of the former, which can be very entertaining, require a degree of internal consistency, but don't need lip service to an at least vaguely plausible extrapolation of what we may regard as 'reality'. Science fiction lives more in the realm of 'what if?'; whilst almost all is meant to entertain, it also can encompass prophetic insight.
A 'Society of Fantasy Writers' could easily fabricate the scenario outlined in this article.
Science fiction writers ought to be more wary; the better among them could construct a highly plausible future in which ideas are no longer 'property'; a future with free for all 'derivation'; an immediate cultural renaissance; a time when the truly creative are acknowledged not so much for what they have already written, but for how that written presages more to come of similar quality. Little imagination is required to foresee a societal shift towards 'patronage' funding, in anticipation of more to come, of those whose work is appreciated, and consequent collapse of parasitical publishing empires, along with numerous middlemen, which collectively arrogate the lions share of income generated through vending creative works.
No, it's better. It's a future where rich people who want to pay so that everyone else gets great work for free just pop out of nowhere. Surely, if those people are there, they'll also be generous as well. If we're leaving reality, why not paint the best utopia we can come up with? The writer of that proposal certainly isn't restricted by reality.
These AI generated stories are going to fill the market for unoriginal generic bland crap, and do it with such a volume that consumers will rapidly tire of it. If, however, you as an author can produce genuinely original ideas, formulated into well written, compelling, cleverly plotted stories, and find somewhere to publish them that can have a reputation for human written only, human curated, high quality work, you should be able to outcompete it, albeit with a slight loss of market share to the AI shit amongst those who will read any old rubbish.
And how will you find that place to publish? If people are attracted to a publisher because they have avoided the AI-generated books, then what stops a bunch of authors from sending their AI-generated books to that publisher for consideration because, if they manage to fool them, they'll benefit from the better reputation? Any publisher that is successful at gaining that reputation will have to deal with a large quantity of LLM-written books anyway.
I'm not sure there's anything we can do about this, but I'm certainly not willing to take the destruction of copyright with that as the end benefit.
I write. I have a couple books out there, complete with copyright. I also write fan-fiction, which plays loose with others' copyrights in a more transformative manner, without copyright. I live with a writer who has maybe thirty books out, complete with copyright. We desperately need computers, because terrible things happen to your handwriting after several decades of keyboarding.
A virtuous use for AI would be as a proofreader or a copy-editor. It could warn if I'm using the same simile or word too often. It could probably tell if I'm using the right word out of the set {to, too, two}. It'd hardly be Nirvana, but it could warn us when we're making clangers.
So-called AIs, or any human being for that matter, seeking to access, and perhaps use, information upon which current laws confer proprietorial 'rights', are, if they abide by current rules, customers in a rigged market.
Digitally encoded information is sold as if physical property. Yet, because sequences of digits have no unique physical presence, as would be the pretence when they are inscribed on a physical medium, they inevitably lack scarcity unless kept under lock and key, and chiselled into blocks of stone which require exertion for removal by a thief.
Scarcity is the foundation of market-economics. It has been since the recorded dawn of humanity. That which is scarce can be bartered for other things. Eventually, conveniently portable scarce objects (e.g. gold coinage) made trading more flexible. Nowadays, we use pieces of paper (or digital abstractions therefrom) in which we place trust that other people too will regard them as acceptable alternatives to gold, silver, etc.
The scarcity of supply (or of ease of access) for goods and services, coupled with demand and competition in supply, determine the current price.
Things lacking inherent scarcity, e.g. salt (sodium chloride crystals) found in huge deposits, as in the English county of Cheshire, are marketable too. The (retail) vendor is offering the considerable convenience of the product having being dug out, cleaned, and packaged, over the hassle of searching directly for deposits. Because what is in the packet has little intrinsic monetary worth, competition centres upon presentation, and the overall price remains very low.
Digital sequences have zero intrinsic scarcity but, just like table salt, they can be 'packaged' in various manners e.g. curated collections, and reliable Internet access (including streaming). However, attempts to put a price on individual sequences, no matter the cost of their production, are spurious. Only the artifice of imposed supposed scarcity via legally enforced monopoly rights can make a basis for 'price'. Yet, in the total absence of competition in providing any particular sequence, prices are arbitrary; they are determined by what 'the market can bear' (i.e. what people are willing to pay) instead of 'price discovery' in an ethos of competition.
In other contexts of markets, monopoly (and monopsony) powers are deemed unacceptable. Nations have 'antitrust laws'. However, in this Neoliberal era of broken markets and of conglomeration, law of that nature is little enforced.
Transition from the analogue to digital era has occasioned considerable worry among people anachronistically claiming ideas (however incarnated) as property like oxen, asses, and motorcars. Initially, concern arose over the increasing ease with which information expressed in analogue form could be copied (e.g. home taping and early varieties of photocopying). Full-blown domestic access to writable digital storage devices, and later ease of sharing 'content' via the Internet, were the beginning of the inevitable demise of digital 'rentier' economics.
AI's voracious appetite for digitally expressed information cannot be quelled/regulated globally. There shall always be Internet connected AIs accessible to the public, many of these being beyond the jurisdictions of IP rentiers. Perhaps, AI is the final straw to break the back of ersatz scarcity and the attendant sense of entitlement it arouses.
Whine and whinge as they might, copyright rentiers shall soon be powerless to stultify creativity and access to knowledge.
Scarcity is the foundation of market-economics. .... Long John Silver
Indeed, and hence the roaringly rewarding expensive trade and devastatingly crushing losses to made in both either the sharing or not sharing of such Top Secrets and Sensitive Compartment Information which interested and terminally invested third parties might prefer to not see the present light of day in the full glare of publicity following on after a 0day trading event ...... for that markets economy foundation can instantly be turned to COSMIC* dust whenever either both too much and/or just enough is made known about the worlds and realities which abound and surround you .... and can read and lead you.
COSMIC* .. Control Of Secret Materiel in an Internetional Command
If we decide that modern AI is useful, then that's exactly the situation we're in. My view is that it is not sufficiently useful to justify exempting training anyway, and that even if it was, the companies that intend to make a lot of money from the result of the training can pay for their training data. We usually don't get exemptions for being useful. No matter how useful the code I write, I don't get my requirements for free. AI companies don't deserve an exception.
People complain about AI being biased due to the biased information it gets trained on, yet here we are people actually *DEMANDING* that AI be biased by limiting what it is trained on. If AI is only trained on content that people have chosen to sell it, then it will be biased to content that people have chosen to sell it.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
