I keep thinking that copyright lasts too long. It seems to me like any regulation around AI training is most likely to benefit a small handful of giant corporations, which will be the only ones that can afford to license content and train newer models. And the small handful of mega corporations that own the rights to most modern media will get their payday. In the end culture ends up being owned and controlled by a few big corporate interests.
I'm sympathetic to artists that want to exclude their works from giant training datasets that primarily end up benefiting the big players in AI without giving anything back. In a way, the first big digital data heist on mankind was executed by social media. Founders of various social media sites became extremely rich thanks to regular people's content being posted and shared (often illegally), without giving anything back to creators.
We've already reached peak data and high quality datasets provide phenomal performance improvements. The tech itself will push for gatekeepers that create the mountains of training data required.
It's amusing that they actually use the word guild. If the medieval guilds had their way London would still only have one bridge. The irony here is that creatives have an advantage that workers in every other industry lack. Humans value and will continue to value artisinally produced goods. Creatives will always have that in a way that 18th Century weavers and spinners did not. That appreciation also leads to the creative industries being subsidised essentially everywhere in a way that is likely to increase over time rather than decreasing.
Will the rise of AI reduce opportunities for creatives? Almost certainly, but unlike essentially all other industries, it won't be wiped out by the rise of automation because humans won't stop liking things that are created by humans. As in music, there will be a shift to performance where customers / clients are engaged in the creative process. In many ways, it will be a return to something like the golden age of portraiture as people pay for engagement.
There are huge opportunities for creatives in the age of AI to create new art forms, created in new ways for new kinds of consumer. Creatives can choose to engage with that or to throw sabots. As London's liverymen show, guilds cannot stop the tide, the opportunity is to become something new that floats on the rising waters.
This. Entertainment is not a zero-sum game, people are quite happy to enjoy two great works rather than one. The margins of profit might be lowered, but so too will be the costs of production in AI.
> That appreciation also leads to the creative industries being subsidised essentially everywhere in a way that is likely to increase over time rather than decreasing.
Not only had it been decreasing for some time now, people are excited about making good-enough music in AI generators. I'm even surprised by your use of "subsidised" rather then "paid for".
We had a chance for the sponsors during the best times of Patreon. Now it's just going to get worse.
if it weren't from guilds all towers and bridges would have collapsed and Maritime trade world never have happened.
then, many centuries later, and land owner elite capture, the guilds become in England just an extension of the peerage system and then it becomes what that phrase above describes. btw most medical labour organizations today work like that protectionism scheme and nobody complain much.
It might well be neoliberal dogma but the rose that got walked through London last week by the Watermen is a reminder that it is also true.
People should complain much more about doctors’ guilds. There are limits on how many doctors can be trained in the US and the U.K. largely as a result of lobbying by doctors’ guilds. They’re called professional associations these days but let’s not pretend they confine themselves to enforcing professional standards.
1. Create a new type of copyright called "training right".
2. This "training right" also applies to names.
3. This "training right" also applies to all usage with AI, even if it doesn't involve training (e.g. as an input to the AI software, such as Img2Img).
4. All AI-generated materials to be captioned as such, and all their activities catalogued and logged.
5. Public domain is no longer public domain (public domain has no "training rights" by default), and freely licensed media is no longer freely licensed (freely licensed media has no "training rights" by default) because "it would not have been possible to foresee its use in a dataset to train an AI model".
This collective of stooges of the European content industry seems to be somewhat irrelevant. Never even heard of one of their "prominent" supporters. This seems like one of the thousand attempts to protect the very mediocre European domestic content industry. As a German, I have seen so many of such initiatives, often even funded by universities or state near bodies.
In music, fine arts, video games, photography there are tons of extremely successful European content producers. Cinema/TV are also quite big too (in no small part thanks to EU legislation incentivising it, but it's top notch quality so who cares, win-win-win).
Mediocre European domestic content industry? You must be thinking of a very specific niche, when it comes to fine arts and music Europe has been a leader since ever.
Unfortunately those big and famous European artists are all dead for decades/centuries. Kinda difficult to monetize today.
Not that we should enable that. Mozart, Beethoven, Vivaldi and Bach are cool, but who should be making money now from their IP? I'd say nobody, it should already be part of CC IMHO.
> Never even heard of one of their "prominent" supporters
no surprise here, you don't seem the type to appreciate art, in any form.
I guess you don't even watch Netflix...
> very mediocre European domestic content industry
Or maybe this is an attempt by artists to protect their works from being exploited by big corporations in exchange for nothing. Which is fine to me, even if their work really were mediocre, nobody can deprive authors of their rights.
Thanks for actually reading the article, and yes, my opinion is kinda similar here
It will be the same story as with Spotify, they will be happy to license it for a bit company for pennies which then again will license it upstream for a big amount
what the fresh hell is this now? What kind of unrealistic and dumb regulations will they now release with yet another costly initiative that European taxpayers don't even want or know exist
> Problem is that, such regulations give ideas to other governments
That because Europe has always been the front runner when it's about advancing the human species and where the best ideas, sooner or later, always win.
This talks about licenses and stuff, but it seems to forget what the current status quo is.
> People: I don't want my stuff to be used to train models.
> Companies: To use our service you grant us perpetual license to use your stuff in whatever way we want, and also the right to sublicense so we can sell your stuff to others while granting them the same rights.
> People: Sure, here you go! Here's my art/code/voice/face/photos/videos/telemetry!
> Companies: [use data according to the license that was granted to them]
> People: pikachu_face.jpg
So enforcing what the manifesto wants to enforce wouldn't change much, if anything at all.
Disclaimer: I have pirated others' stuff (e.g. anime, manga, novels, music, I have shared memes with others (that's distribution), etc), so I can't complain when others pirate my stuff without being a hypocrite. The most I can do myself is call them out for profiting off it.
Companies already are doing what they say they are doing, even if it's unenforceable. The profit from doing so is more than the liabilities from any lawsuit might be receiving.
History shows that creating more legalese text won't add any more teeth. A specific example is how 90% of websites[1] are in violation[2] of the GDPR already.
The manifesto raises a good point about data not protected by copyright; but one can argue that models created from copyrighted content, what most artists I see are complaining about, are already in violation of copyright due to being derivative works (IANAL), yet already-existing copyright laws aren't helping[3].
So they way I understand this manifesto is, they are not adding anything new for copyrighted works (AFAIK, IANAL), but they are taking away rights from public domain works (let's see how that plays out with Disney, for example).
Hopefully this additional context clarifies where I'm coming from.
[1]: I'm making up that number but I'm probably lowballing it anyway.
[3]: But I bet if someone made a model from TV shows, now suddenly the already-existing copyright law would magically work for those specific cases. So basically, these randos without power are making a law that will only be enforced by those with power.
Oof my bad. I should have added "websites I visit", or "website I've seen". I don't keep track of them, but if I try to think of any websites that I've visited e.g. since last month that are not in violation, only ~4 come to mind. And since I don't have any exact numbers, I have no choice but to give a conservative estimate.
I'm usually careful with wording since not everyone tries to read in good faith, but I was probably too tired at the time to notice.
I apologize for any offense that my comment might have caused.
No offense, I simply wanted to show that, even if what you say it's true, there is no "police state" trying to monitor and eventually shut down the Internet, like many are suggesting here and that there is a lot of leeway for the smaller actors.
Nobody is really worried if the website for Go packages doesn't fully respect GDPR, except maybe Go it's a Google product, but the best they can infer it's the demographics of Go developers, which they probably already know.
What is important is that the data collectors can be held responsible for the infringements.
Same objective I imagine for EIGAR (I know some of the founders), it is not to stop generative art using AI, but to stop pretending that massive data scraping of copyrighted material is fair use when Disney or Amazon or Netflix or any other tech mogul profits from them.
Hollywood writers and the actors' guild have already gone on strike over this, this is nothing new or unexpected, it was simply waiting to happen, it's Newton's third law of motion applied to humans.
Since this is an association of creatives that will be left behind anyway, even if all AI-companies were european, that comment doesn't really make sense. The american SAG-AFTRA also protested some usage of AI and you wouldn't really say that Hollywood Actors wanna be left behind - quite the opposite actually.
it is no different from what we have right now. With big tech you will just have more opportunities to do something yourself rather than paying for "hollywood papa" or bend the knee to disney.
that's the wrong question, it is like asking "how many peasants are there nowadays"?
of course there are many more people that do not need to plow the fields all day to provide for their 11 children and that can entertain themselves with arts, that's not the point.
the point is that mass manufacturing easily reproducible standardized artifacts becomes industrial production, which has nothing to do with art.
I'm sympathetic to artists that want to exclude their works from giant training datasets that primarily end up benefiting the big players in AI without giving anything back. In a way, the first big digital data heist on mankind was executed by social media. Founders of various social media sites became extremely rich thanks to regular people's content being posted and shared (often illegally), without giving anything back to creators.