The entire AI community is in complete denial regarding AI models that violate copyright laws.

The majority of the AI community will say that AI models that violate copyright laws are a looming, but not yet present "concern". In reality, these models have already been out for awhile now, and are being widely used all over the internet.<br><br>I realize that this type of content "violates the rules" of this chamber, but given the sheer looming threat it presents to the content creation industry, I think it's crucial to comment on.<br><br>I am not posting this so that you can personally go and use these models to violate copyright laws. My goal here is to inform you of this problem, so that you can then comment on it in a meaningful way.<br><br>The models I am referring to are not "AI generated" in the classical sense. Instead, these models use large amounts of training data and then proceed to memorize this training data. If you input the right prompt in the right way, the model can be tricked into "regurgitating" copyrighted content. <br><br>These models do not exist in some obscure corner of the internet. Rather, they are widely used by the entire AI community. I personally know multiple small businesses that are using these models to violate copyright laws (for example by regurgitating children's TV shows and then selling the result on YouTube for a profit).<br><br>For example, "borderline illegal" content like AI generated Rick and Morty scenes with voices "similar" to the show are all over the place. <br><br>And then there is the actual hardcore illegal content like AI generated episodes of Spider-Man: The Animated Series (www.youtube.com/watch?v=KvvdJ5TwU_s) or AI generated episodes of The Dragon Prince (www.youtube.com/watch?v=l3D9Y6qBnVM).<br><br>I've seen models that can regurgitate extremely accurate scenes from Avatar: The Last Airbender, The Legend of Korra, Batman Beyond, and many many more.<br><br>I am not posting this so that you can go and personally use these models to violate copyright laws. And I am not posting this to "slander" the AI community or to exaggerate the problem.<br><br>This is a very real problem that the content creation industry will have to deal with.<br><br>And given the fact that this is a looming threat to the content creation industry, I think it needs to be discussed in a far more open and honest way, rather than the entire AI community being in denial about it.<br><br>My personal opinion is that these models need to be illegal in the same way that torrenting copyrighted material is illegal.<br><br>In addition to this, the regulatory landscape surrounding these models needs to be entirely reworked around the idea that these types of models are already widely used all over the internet.<br><br>The current regulatory landscape pretends that these models are some sort of fringe, edge case that might become more popular in the future. In reality, these models already exist, are widely used, and are already a massive threat to the content creation industry.<br><br>I think that the AI community needs a more nuanced discussion around the severity of the problem, the severity of the threat, and what should be done to combat this threat.