- cross-posted to:
- technology
- cross-posted to:
- technology
You must log in or register to comment.
I’m starting to think commercial AI should be banned. if the only way to make useful models is by ingesting human culture, then all humans should benefit from it without having to pay to have that culture shat back out in response to a prompt.
Im starting to agree with that premise. Since these models only exist using the public’s data they should be public models only. No commercial use.
The “issue” is that this logic applies to all human creations as well.
i disagree. IP laws have more or less handled humans stealing ideas from humans for commercial gain. not perfectly by any means… but both the scale an impunity and frankly the entitlement exhibited by these GenAI companies is on another level.
no matter how many times people make the argument that AIs are just “doing what humans do”, it fails to sway me. an AI copying, ingesting and tokenizing other people’s intellectual property is nothing like a human watching a video or hearing a song and creating something based upon or derived from it. a database backed algorithm does nothing even remotely like a human mind. it’s using software to process and regurgitate the works of others, and that is pretty plainly IP theft.
I’m not saying the process is exactly the same but conceptually it’s quite similar. Humans don’t create original ideas. They build on what came before. Maybe a truly brilliant artist or inventor adds 1% new ideas. That’s not enough to justify the extremely broad ownership of ideas that exists in our society. These laws implicitly assume that ideas were created from nothing through the sheer brilliance of the creator. Pure nonsense.
Humans have been freely copying each other for millions of years. It’s how we built everything we have. Ideas and art were not meant to be owned. The very concept of owning something non-physical is violent and authoritarian in nature. Without physical possession, the only way IP laws can be enforced is a global police empire, which the US has successfully created for its own enrichment at the expense of the global poor.
So in that context, the fact that AI is borrowing human ideas and then profiting from it doesn’t bother me any more than that humans do the same thing.
Humans have been freely copying each other for millions of years.
False. Master artisans have been keeping their knowledge secret in order to maintain a competitive advantage, only eventually passing their knowledge to the most advanced of their apprentices. Tons of knowledge has been lost over the millenia to Masters taking their knowledge with them. Temporary monopolies (Patents) and Copyright protections, in exchange for making the knowledge public, is what has enabled its exponential expansion.
Keyword being “temporary”. We have Disney to thank for turning Copyright’s temporariness into a mockery of itself.
I’m not sure I buy this argument but I will admit that I’m not extremely familiar with the hoarding or sharing of trade secrets prior to patents. Any recommended reading on this topic? If your logic is correct, patents should be as short as practically possible to encourage information sharing.
I don’t see how this applies to copyright though. Are you concerned people will create works and then bury them? I don’t see the risk here.
On patents and lost knowledge:
https://ipwatchdog.com/2022/09/07/history-patents-can-teach-us-world-without-might-like/id=151264/
https://en.m.wikipedia.org/wiki/Stradivarius
https://en.m.wikipedia.org/wiki/Artificio_de_Juanelo
If your logic is correct, patents should be as short as practically possible to encourage information sharing.
Modern patents require the disclosure of information prior to being granted, so anyone can access the knowledge to build on it from the start. The patent owner’s rights are enforced after the fact, by punishing anyone who tries to make money off the invention without a license from the owner. Their term is generally reasonable for mechanical inventions, with a maximum of 20 years, and the cost of maintaining the patent grows exponentially. Main problems are the term, and whether a patent should be granted at all, when applied to non-mechanical items, like software, medicines, organisms, etc. which don’t follow the same pattern of investment vs. incentive.
Copyright, was initially intended to let publishers have some time to get their investment back, between printing, distributing, and selling copies of a book. Initially, in the 18th century, that was set to 28 years. However, instead of staying true to that intention and adapting to new forms of distribution, with the internet being the latest one, Disney lobbied like crazy to get to the current “until author’s death + 70 years” term:
https://en.m.wikipedia.org/wiki/Copyright_Term_Extension_Act
That, is a complete mockery of the initial rationale. With digital distribution, Copyright should’ve shrunk to a fraction of the original 28 years, not grow even longer!
Are you concerned people will create works and then bury them?
The concern is that people, particularly publishers paying an advance to an author, would not want to do that if they didn’t have some assurance on the return of their investment. Nowadays, something like 1 or 2 years after publication, would be more than enough, even for films, which get most of their revenue during the first few weeks after release. Games follow a similar pattern, when they don’t require an ongoing subscription.
Injecting logic, facts, and moderation into an AI conversation? Get the fuck outta here!
IP laws have more or less handled humans stealing ideas from humans for commercial gain. not perfectly by any means…
“Until author’s death + 70 years”… not perfectly, is WAY of an understatement.
an AI copying, ingesting and tokenizing other people’s intellectual property is nothing like a human watching a video or hearing a song and creating something based upon or derived from it. a database backed algorithm does nothing even remotely like a human mind. it’s using software to process and regurgitate the works of others, and that is pretty plainly IP theft.
Wrong. There is no database of the training data in the model that “regurgitates” abstracted concepts from it.
Sure, but the argument isn’t “should we ban work that is based on the study of past cultural creation” it’s “we should prevent computational/corporate exploitation of past cultural creation in order to protect the interests of humans.”
I’m saying we’ve already allowed corporate exploitation of human culture for centuries. But yes, by all means, if AI is the last straw then I’m with you. But I want people to see the broader picture and not hyperfocus only on AI.
SHOW ME WHAT YOU GOT
I think the alternative: copyright should be looser. It usually only benefits corporations and lawyers.
Though it would be naive to consider AI companies and ally in a goal to reduce copyright terms.
Same can be said of commercial Schools, Colleges, and Universities.
Companies use pirated content: line go up! 📈
Everyday people use pirated content: STRAIGHT TO JAIL!
If companies were really treated like people, they’d be in jail right now - at least in the US. But they wouldn’t come out reformed, just beaten and bruised, ready to commit more crimes.
The point of treating companies like people is so no one in those companies can be held accountable. The worst case for them that the intangible “coorporation” did something wrong and now it has to go away, so the entire board moves to a new company under a new name that owns the same properties and has the same practices. Only now they have practice obfuscating their crimes.
What ends lives and careers for people are just a minor inconvinience to coorporations.
The CEO of a corporation should be the living embodiment of that corp. Kind of like Subway in Community
The funny thing is that Google can’t make a big fuss about it without screwing themselves when they are inevitably sued for the same thing.
Surprise!
Hehe we do a little scraping
I am shocked that an AI company would do this. This is unprecedented.
It should be assumed by default that data collection is almost always non-consensual.
