Japanese court rules that AI cannot be issued patents
japannews.yomiuri.co.jp
japannews.yomiuri.co.jp
I can't pretend to be a Japanese legal scholar but at the end of the day both an AI and corporation are just property.
Can't you just put a thin legal wrapper around your AI and call it invention-station Inc or something and thus christen it with all the rights of corporations?
If a company can own a computer and it can own software then it can certainly own what's produced by the software on the computer...
I don't see how this in practice changes anything. It sounds like you can still do the thing with slightly different paperwork.
I think they may be the first to suggest in court that a different IP system might be used for AIs though?
A human can also reproduce very realistic copies of things and can even pass things off as the "genuine" thing.
We already have copyright law and counterfeit law for this.
Even if AI doesn't work the same way as parts of the human brain, I don't understand what difference it makes, whether AI was involved in the production of work legally, morally, or philosophically.
I'm stating plainly: I don't see what differences between AI and human brains are relevant to the enforcement of copyright.
What is fundamentally changed by an LLM with respect to these three thought frameworks?
I would also argue, that models learned from publicly available data need to stay public as well.
Training is a high level word, and because we engineered LLMs we know more specifics about how their training works, so when used in context with LLM, it can mean something more specific.
I don't know the full mechanics of how training/learning works in the human brain, but it's I've been studying casually for 30 years. It's possible that parts of our brain work very similarly.
What I'm trying to say is, regardless of underlying mechanics, I don't understand why using copyrighted works for training (anything) is a against some legal or moral framework.
-> I don't know how human thinking works, and according to my research, nobody does (not yet at least).
It might be exactly the same, or it might be similar, and it might not. This is all irrelevant to what I am saying.
I am saying, it doesn't make any difference according to my understanding of existing law.
I also don't understand what moral issues there are here that are specific to LLMs, and apply only to LLMs.
When I say that LLMs train, and humans train, I mean from a systems perspective, they both use the input of the data for similar purposes.
I ask the question: Why is one different than the other? Why are humans allowed to analyze and learn from art, but an LLM isn't? If they are different, and not analogous, that's one possiblity... but still doesn't answer my question: why does this unspecified and unknown "difference" matter?
Fair use is a thing.
Also you seems to misunderstand something, the article never stated you couldn't patent an AI output, simply that the AI itself cannot be the author.
It does.
Oracle vs Google, Google copied Oracle's API, is considered fair use as last judgment.
Music sampling is also fair use. Very popular and lucrative music's used samples of copyrighted material.
Almost nobody ask for permission.
I'd also like to point out that you said
> Fair use laws, if the country has them, don’t include profit.
And now you start to add a lot of conditions depending of that. Oracle vs Google isn't a "false equivalency" it's litteraly a fair use case, where the one who copied the original work did it for a commercial product.
The article state that AI cannot be the owner of a patent, it have to be a human.
https://en.wikipedia.org/wiki/The_Animatrix#The_Second_Renai...