• This is ridiculous, how would you even determine that a given image was created or edited with AI? And what about images that were generated with Photoshop, do they also require a watermark?

    • I think that is an enforcement problem. The law is intended to stop people from doing this. If examples are found the government now has a reason to hunt them down and a punishment to discipline them with.

      I’m sure that most violations of most laws aren’t caught. That doesn’t make the law useless.

    • It is an interesting question of what constitutes human art compared to machine generated art. Ultimately it is the human who guides the creation process and curates the final image using their aesthetic. The process doesn’t seem that different from photography. And as you note, it’s not clear where you draw the line in terms of computer assisted art. Drawing using Photoshop takes a lot less training than oil painting for example, yet we don’t see digital art as being a lesser medium.

      All that said, I imagine the area that China is concerned with would be AI generated content passed off as news. You can easily generate deepfake video of a politician for example, and having rules to prevent such a video being passed around as real seems prudent.

      • All that said, I imagine the area that China is concerned with would be AI generated content passed off as news. You can easily generate deepfake video of a politician for example, and having rules to prevent such a video being passed around as real seems prudent.

        The solution to that is verifying sources and cross-referencing to make sure that its actually real. It has been possible for a very long time to edit images and videos in a way that appears real, AI just makes the process faster.

        • I think the speed does make a qualitative difference though. With AI it’s now possible to churn out content very quickly and very cheaply. So, having a way to track factual content is becoming increasingly important. I do agree that focusing on watermarking factual footage to make it verifiable instead of AI generated content would probably be more productive.

          • Watermarking factual footage in a way that people can verify metadata would also be an interesting tool for journalism to be able to focus on evidence to improve confidence of facts. Taiwan fights misinformation by having easily available tools people can use to verify information and focus on improving critical thinking of the people. fighting the infodemic and pandemic without takedowns and lockdowns seems to be the best approach. Tools that prevent people getting tricked by disinformation and verifying information for the infoedmic and improving clean air with things like CR boxes and c02 sensors and clean air regulation for the pandemic.

        • Sure, but the question is whether it’s the technical skill that’s important or the vision the artist is trying to convey. In my opinion, AI medium allows people to convey their ideas without having the technical skills to do so. So, it’s just opening up the visual medium for a broader segment of the population who have ideas in their heads that they want to share with others.

    • If you look into this decision, it’s more that China is worried about deepfakes, which is a very real concern.

      In recent years, deep synthesis technology has developed rapidly. While serving user needs and improving user experience, it has also been used by some unscrupulous people to produce, copy, publish, and disseminate illegal and harmful information, to slander and belittle others’ reputation and honor, and to counterfeit others’ identities. Committing fraud, etc., affects the order of communication and social order, damages the legitimate rights and interests of the people, and endangers national security and social stability.

      This is likely easy to enforce at the model level, if you have a model that generates lifelike impressions of real people. Enforcing it per image would be impossible I think.

      But there are people celebrating this like it’s some luddite attempt of China to hold back technological progress for the narrow aim of protecting IP. Any “communist” that is disposed this way, read the quote above a few more times. When the sewing machine was invented, did we hold back the sewing machine so that more tailors could keep their jobs? Why should it be any different for “artists”? Is the solution to alienation turning back society to the dark age? Or is there already a theory of revolutionary change that venerates the acceleration of revolutions in the forces of production? If you hate AI art you are a reactionary.

        • Property is a spook. Post scarcity is here. We have to adapt our economic system to the new reality.

          Shakespeare didn’t invent Romeo and Juliet you know? Back in the day artists didn’t have a concept of IP. People would just freely rip off and adapt one another.

        • Why? IP is a bankrupt concept. Instead of protecting IP on behalf of artists, we should abolish IP for AI companies. Enforce open sourcing of AI models that use other people’s data, nationalize big tech, so on

  • For the comments that say “what constitutes as AI art” well, if you have decided to take a look at any AI generated art you’d easily be able to tell. And no matter how many art styles it can imitate, it’s exactly that an imitation you can see the differences and the issues that come with trying to do that.

    For anyone saying this takes less skill… No it doesn’t, you still need to know exactly how to work with the AI to not make its generated images have broken anatomy, melted skin and broken art.

    And for those that say this is horrible because “How dare China make a law to regulate AI art from taking away money from real artists”… Hear yourself and see where these technologies have gotten us with Facebook, Twitter, Youtube and so many privacy-invading tools growing unimpeded.