Obviously fuck the capitalists and AI scammers. But reading and learning from a library, then writing and selling your own book based on that is NOT stealing. It’s the wrong argument.
The answer obviously is to keep the actual source material and libraries and book archives open and just run open source AI models at home. You can run smaller versions on a solar powered PC no problem.
The issue with trying to make “AI is just stolen” happen is that it will make open source AI models illegal. AI companies would love that because they can afford to license and pay or work around or obscure or whatever. The “intellectual property” argument is always a disgusting capitalist one. Knowledge is either free or nothing is.
Every book ever published, every article anybody ever wrote, every comment anyone ever posted on a public internet is “consent” to read and learn from it.
I’m pro piracy as you can tell. The idea that something can be out there publicly on the internet but it’s “not consent” to read is the intellectual property one. Look at how they try to gatekeep publicly owned scientific papers. Big AI is clearly hypocritical doing this, but corporations are just soulless, amoral programs executed by sentient humans.
But the RESULT of all that (e.g. deepseek) should belong to all people. And THAT is why these IP arguments by fuckAI are dangerous, because it is only a threat to open models. The answer is open source (or weight) AI models and with advances in computing to run them locally.
Well yeah, they suck. I’m a fan of anna’s archive.
But this doesn’t change the fact that AI models will continue to improve, and the tactical question is if we give them munition to monopolize it using “intellectual property” rights. I want open source/weights models to use locally without paying some license to meta or reddit or some publisher cartel.
But reading and learning from a library, then writing and selling your own book based on that is NOT stealing.
It could be argued that these AIs are not actually learning but collecting and rearranging. That’s still stealing in my book, especially if it happens on a massive industrial scale and by a megacorp instead of a person.
That is incorrect though, it follows the fallacy that it’s just like a big database where all that (much larger) data is being copied and compressed into. It’s called machine learning and denying the reality of how it works is just not useful.
Imagine you study as an engineer in whatever field, but now laws have been passed that you only licensed the knowledge from university and publishers. If you work you have to demonstrate who you learned it from and then pay royalty fees. Obviously that would be insane for humans, but I do forsee that they will try to do this for machine learning. Because of the argument you made.
So any open source / weight model you find and could run locally (like e.g. deepseek) will now be illegal because you can’t prove where “dey tuerg dur dartae” from.
Thus all potential future gains from AI will be monopolized, while the costs socialized.
Obviously fuck the capitalists and AI scammers. But reading and learning from a library, then writing and selling your own book based on that is NOT stealing. It’s the wrong argument.
The answer obviously is to keep the actual source material and libraries and book archives open and just run open source AI models at home. You can run smaller versions on a solar powered PC no problem.
The issue with trying to make “AI is just stolen” happen is that it will make open source AI models illegal. AI companies would love that because they can afford to license and pay or work around or obscure or whatever. The “intellectual property” argument is always a disgusting capitalist one. Knowledge is either free or nothing is.
Large amounts of data were pirated, which is apparently legal if you’re training an AI. Source: Meta lawsuit
Library implies consent that was not there and access to the public that was not there.
Every book ever published, every article anybody ever wrote, every comment anyone ever posted on a public internet is “consent” to read and learn from it.
I’m pro piracy as you can tell. The idea that something can be out there publicly on the internet but it’s “not consent” to read is the intellectual property one. Look at how they try to gatekeep publicly owned scientific papers. Big AI is clearly hypocritical doing this, but corporations are just soulless, amoral programs executed by sentient humans.
But the RESULT of all that (e.g. deepseek) should belong to all people. And THAT is why these IP arguments by fuckAI are dangerous, because it is only a threat to open models. The answer is open source (or weight) AI models and with advances in computing to run them locally.
Except these people are turning around and burningdown the libraries once they’ve read all the books.
Well yeah, they suck. I’m a fan of anna’s archive.
But this doesn’t change the fact that AI models will continue to improve, and the tactical question is if we give them munition to monopolize it using “intellectual property” rights. I want open source/weights models to use locally without paying some license to meta or reddit or some publisher cartel.
A powerful argument if one cannot tell the difference between a person and a product.
It could be argued that these AIs are not actually learning but collecting and rearranging. That’s still stealing in my book, especially if it happens on a massive industrial scale and by a megacorp instead of a person.
That is incorrect though, it follows the fallacy that it’s just like a big database where all that (much larger) data is being copied and compressed into. It’s called machine learning and denying the reality of how it works is just not useful.
Imagine you study as an engineer in whatever field, but now laws have been passed that you only licensed the knowledge from university and publishers. If you work you have to demonstrate who you learned it from and then pay royalty fees. Obviously that would be insane for humans, but I do forsee that they will try to do this for machine learning. Because of the argument you made.
So any open source / weight model you find and could run locally (like e.g. deepseek) will now be illegal because you can’t prove where “dey tuerg dur dartae” from.
Thus all potential future gains from AI will be monopolized, while the costs socialized.