More and more images on the internet are created by AI image generators. Some may have heard of AIs like DALL-E, Craiyon, Artbreeder or Midjourney. This system are enjoying increasing popularity on the internet. However, online art platforms are struggling with the AI-generated images. Now the first ones are reacting with bans.
As AI image generators have become more popular, art platforms face new challenges. Many are confronted with a flood of images behind which there are no human artists. With artificial intelligence (AI), practically everyone on the Internet can now have images generated quickly and easily.
First upload bans for AI artworks
To counteract this flood of images, art platforms such as Fur Affinity, Newsground and InkBlot have issued a ban. Fur Affinity has its on September 5th guidelines modified for uploading images and justified the step with the approach of AI image generators like DALL-E 2. For the digital art platform, an AI-generated image is without artistic value because it would only use the works of thousands of artists. An image created in this way is therefore only a derivative.
Also inkblot announced on Twitter a zero-tolerance policy for AI images. As a result of this announcement renewed inkblot its content guidelines and announced that it would remove AI artworks immediately. The process of machine learning from other artworks would violate applicable copyright on the platform. The platform explicitly includes popular AI image generators such as Midjourney, DALL-E and Craiyon in the ban, but also wants to take action against other generators.
You may also be interested in: “In 2062, machines will be as smart as humans”
Does AI copy the works of human artists?
It remains to be seen whether other platforms will follow suit. The frustration of some users is vented in forums and on Twitter, who remark that human works of art are lost in the mass of artificially created works. In addition, the “living” artists would slowly be deprived of their livelihood. Because AI changes original works only minimally and thus interferes with the copyright of the artists.
Some generators want the generated images to look similar. DALL-E 2 offers the function of creating images from a template. In this way, a new combination of styles can arise from an input text phrase in the text-to-image generator. In addition, the AI can add or remove objects in images. This looks amazingly realistic making creating plagiarism much easier. The slight modification of a work by AI poses new challenges for copyright law.
The end of art?
Distinguishing an AI-generated image from an original will only become more difficult in the future – mainly because it takes only a few seconds to create and upload. In addition, the way AI works is in most cases unknown to the public. Only the developers themselves know which image databases their AI is accessing. Exceptions are open source projects, which are accessible to everyone.
But of course human artists are also inspired by the works of art of others. It is not always easy to recognize when it is plagiarism and when it is not. However, the speed with which AIs can access original works and generate new images from them greatly exacerbates this problem. For machine learning (ML), artificial intelligence is trained with millions of images and image descriptions. DALL-E 2 KI about can on 650 million images access, which can be accessed via the Internet.
Whether a ban on AI-generated images on art platforms will solve the copyright problem is uncertain. The social discussion about AI art is only just beginning.