@jeff@McMongoose@Shadowman311 Using anyone's likeness is already illegal due to the Crispin Glover Back to the Future lawsuit. It wouldn't take much to argue that basing AI on someone's voice is stealing.
The bigger question is if something that "sounds like" someone coincidentally is infringement. If a jury decides it "sounds like" something it becomes infringement.
"AI not copyright protected" ends with human editing. If Disney decided to just use AI for everything it's kinda hard to fight it because there's millions of excuses and examples to draw from. If the Eiffel tower can be copyrighted by adding a lightbulb AI is copyright protected when you add a dot on the work. :back_from_gab:
VAs fucked themselves though. I remember when they were making the union a few years ago one of the rules they had were that if you hired a union VA you couldn't use non-union VAs. So you're left with low quality amateurs, or high quality retards with low quality faggot VAs.
Voice AI solves the problem because you can circumvent the VAs. You can make them sound like a known VA or have a better computer voice for your voice acting. As long as you don't make the voice sound like a known person you'll have no legal issues.
VAs can't even sell voice banks because they may ban that stuff by the union. Imagine, a mark hamill voice bank license could make you infinite money for decades but they would shoot themselves in the foot and ban this because they're retarded. :back_from_gab:
Not true. Learn copyright law. That it's produced by AI does not mean it's not a "derivative work." You need more than "it's heckin' produced by AI" to win the day.
@Humpleupagus@McMongoose@zemichi@Shadowman311 if a work is 100% AI generated it is public domain. this is for when you build atop existing created works using AI or any other mixed attribution work.
Tracing would be the issue for images produced via multiple passess upon passes. This will only be litigated when there's money at issue, as having a market and damages, are common elements in a copyright suit, e.g. point and shoot photos get the least protection, especially where they on a home camera.
Someone would have to argue that image data converted to "data" is a unauthorized use. Or is that Data derivative work itself?
The other issue would be having AI have the ability to reproduce an already existing work. (Like 1:1 or near it) When does that become infringement?
Whatever the case this ends when some autist auto generated a library of AI works passed through 100 times and self references that. What then? :backfromgab:
You're looking in the wrong direction. It's not about how it was produced, it's whether the work is derivative. The dataset used is what matters.
"A derivative work is a work based on or derived from one or more already exist- ing works. Common derivative works include translations, musical arrange- ments, motion picture versions of literary material or plays, art reproductions, abridgments, and condensations of preexisting works."
@Humpleupagus@McMongoose@Shadowman311@zemichi the update states that if something is 100% purely AI generated then it has "no author" and thus exempt from copyright and IP law, effectively akin to public domain.
"A derivative work is a work based on or derived from one or more already exist- ing works. Common derivative works include translations, musical arrange- ments, motion picture versions of literary material or plays, art reproductions, abridgments, and condensations of preexisting works."
@Humpleupagus@McMongoose@zemichi@Shadowman311 you're behind dude. this is very good update to the 19th century style moon runes spoken by old man waynmenth. 100% AI generated would imply there is no clear derivative work made by a human. the images generated by NAI fall under this.
@Humpleupagus@McMongoose@zemichi@Shadowman311 you're looking at old legislation bro. the current working draft is something to read up on. get up to speed and come back to us, thnx.
@Humpleupagus@McMongoose@zemichi@jeff@Shadowman311 That's an interesting argument, because the training data is the actual work, it doesn't matter what it produces, because it is still using the work of the artist. Therefore, it becomes a battle to prove that it is an original work inspired by the artist.
Yes. And it is an issue given the current state of the law given the advancements in technology, much like how the SEC is attempting to use a century's old law to regulate crypto.
It will have to be addressed by the courts unless legislation is first passed.
One obvious issue will be AI that imitates a particular artist, i.e. where it was trained on their works, and that artist has a market.
As the number of works and artists used gets more numerous, it could get more nebulous, though I'm not convinced. A class action may be the proper means of suit... where no one knows for sure who was damaged, but they all know that they all were.
I don't think so either, but to get there, I think legislation will be required. I think that globohomo would lose enough cases where the tech would be in doubt if the current trends continue and the law is not amended.
To your point, I do note the DMCA, which was passed to address copyright liability of providers of internet services and with the goal of allowing the net to grow. Congress recognized that the net was in issue under the then existing law.
It's doubtful that anything adverse to AI progression it's going to happen at this point. Microsoft using GitHub data for OpenAI LLM training really opened the gates. Microsoft knew the data had significant value when they purchased GitHub in 2018.
Even giants like Disney and Getty would struggle to overcome what's already occurred.
They'd have to maintain the training database, probably with logging, and all of the ESI (electronically stored information) rules would come into play.