Whose work whose?

“AI is often seen as a threat to humanism. But we can have both. We can have humanistic interpretations and explanations at the same time. And we can look at art appreciation from the AI side and see what comes out from there.”

Dr Mark Coeckelbergh believes that AI can help us understand art better, with the help of big data. But the worries about using AI cannot be shaken off, as its presence will change many existing conventions and even prevalent laws that will become harder to apply.

Copyright is a good example. Who owns AI-generated art, and how far can one use those images? The current answer is that users of applications such as DALL-E or MidJourney have “certain rights” to the images they ‘create’ with their prompts.

For example, trial users of MidJourney are given rights to use the images as long as they don’t sell or make a profit off them, and give proper credit to MidJourney. If one subscribes to its paid services, MidJourney clarifies that users “basically own all assets they create using Midjourney’s image generation and chat services.” In DALL-E 2, users will receive full usage rights to the images they generate, including rights to reprint, sell, and merchandise.

However, a prompt that allows any user to generate images mimicking the style of any specific artist and subsequently masquerade and profit from them will have copyright laws changing.

The question of artist authenticity is vaguely addressed in how DALL-E 2 advocates for its ability to ‘take an image and create different variations of it, inspired by the original.’

Screenshot from https://openai.com/dall-e-2/

Whose original is the software referencing from?

Most examples of artworks referenced on DALL-E’s website are registered in the public domain, and essentially free from copyright. But how do we know if the data model is learning from only legal, copyright-free artists? After all, the software that generates the images must learn from a big set of images, to react to artist names, styles and imagery prompted.

If the generated artwork mimics a style of a living artist, can one claim Intellectual Property over software-generated work? Will users own rights to the prompts they use and each variation they work on? What happens when another user uses the same prompt?

Dr Mark Coeckelbergh suggests that “many applications have been more like imitating humans, and that's not so interesting. Beyond imitation, there will be creation through pattern recognition, for example. Machines also do things that we don't expect. I believe that there needs to be some human input. It will have to be collaborative between man and machine.”

Artificial Intelligence and image generators are not new. What is new is the surprisingly extensive ability of the software and its accessibility. The need to discuss ways of protecting authenticity is now at the forefront of conversations, as such programs become integral to all forms of creative processes, including art.

We also spoke to Prof. Dr Mark Coeckelbergh who shared some insight on the future and potential of AI. To read the full interview, click here.