For the first time, groups of artists are suing generative AI companies. This attack was motivated by the lack of respect for copyright and the non-existent protections provided for creatives.
Just compared to a few months ago, our way of thinking about images, art and interaction with machines has changed forever: thanks to deep learning models, In fact, Artificial Intelligence is now able to produce compositions based only on textual instructions. And the improvement is continuous: over time the images made by the machines are increasingly realistic; meanwhile, the same AI become viral and widespread (just think of the success of the latest project launched by Binance, Bicasso).
According to the Web 3 community, the AI community is certainly a promising start; but creatives think differently, who are gradually expressing a growing concern about the lack of respect for copyright law. And as protests and perplexities grow, one wonders if the artists who collaborate with AI can really define themselves as such.
The question of consensus
One of the main issues at the heart of the debate has to do with consensus. Artificial intelligence systems are trained from huge amounts of data, which is retrieved from billions upon billions of existing images. These, however, are taken from hundreds of random internet domains, and used without anyone giving their consent. But the reactions of the artists are not slow to arrive, and the first legal disputes open a long clash.
The first alarmisms
Early concerns about consensus management arose at the end of 2022, when experts pointed out that many AI-related issues would require a clear definition by the judicial system. In January 2023, however, the picture quickly deteriorated. For the first time a collective legal action has been opened against three companies that produced generators of art AI: MidJourney, Stability AI (the company of Stable Diffusion) and deviantart.
To move the lawsuit Sarah Andersen, Kelly mckernan and Karla Ortiz, three artists who highlighted that the companies involved violated the rights of millions of creative: were using billions of images available online to train their AI without having had any consent.
The real problem
One of the crux of the matter has to do with the fact that these programs do not generate images based solely on text, but also come to imitate precisely the style of any living artist who, in a few seconds, can thus be completely replaced by machines.
In an editorial released for the New York Times, Sarah Andersen talks about how she felt when she realized that AI could replace her in her work:
"The idea that someone could type my name into a generator and produce an image in my style immediately disturbed me. It wasn’t a human being who created fan art, nor a mischievous troll who copied my style; this was a generator that could spit out different images in seconds".
A copyright infringement
The images that are used freely to train AI are actually copyrighted; according to the plaintiffs and their lawyers, this means that any reproduction of them without permission would be a violation. The companies that use them, in fact, end up benefiting from them commercially, also making great profits.
A damage, therefore, far from hypothetical: for example, it is very likely that AI will end up taking away from artists the classic "commissions"; and in the not too distant future, it is not unlikely that artificial intelligence will end up replacing them completely.
The doctrine of fair use
Supporters and developers of artificial intelligence tools, however, seem to think otherwise. In particular, they defend themselves by referring to the doctrine of fair use, for which it would be possible to use copyrighted material without obtaining permission from the right holder.
As concluded by a spokesperson for Stability AI:
"Anyone who believes that this is not a proper use does not understand the technology and misunderstands the law".
• #Artificial intelligence