Updated: Aug 1
AI art is a rapidly advancing field with the potential for new and groundbreaking creative works at the intersection of human and machine imagination.
One of the main ethical questions surrounding AI-generated art pertains to ownership. AI art raises interesting questions around issues of patent, copyright, and legitimacy. It is not clear who owns the rights to AI-generated art: is it the programmer, the program, or the entity that trained the AI algorithm?
The question is even more pressing when considering the potential for copying or selling artworks created by AI to others for commercial gain. Some would argue that AI-generated art should be free and open source, echoing the "copyleft" or anti-copyright movement. Others argue that AI's ability to create artworks should be considered on an equal footing with human creativity, and that legal rights should be accorded to the generated works. While the assessment of legal ownership is a significant area of debate, the ethics of AI-generated art extend far beyond mere intellectual property rights.
Some artists use AI-generated art for political and social commentary, highlighting the intersections of race, gender, and society. One example of such political art is the work of multimedia artist Trevor Paglen. Paglen's project, "ImageNet Roulette", was created using a database of over 14 million images that had been previously tagged and organized by humans. The artist trained an AI algorithm to randomly select and display images from these datasets, along with keywords that provide insight into the inherent biases in the tagging system. Through this algorithm, Paglen was able to expose the shortcomings in the ways that humans categorize visual data, and the politicized ways in which we label and value the images.
Paglen's project is just one example of how AI-generated art can raise larger ethical questions. Other issues include the use of AI-generated art to create propaganda, to deceive viewers, and to manipulate the public. AI-generated works can also challenge our definitions of creativity, artistry, and authorship. For example, is an AI algorithm that generates images an artist? Is the dataset of images that the algorithm is trained on a collaborator? The role of AI-created artworks can blur the boundaries between authorship, collaboration, and appropriation.
At one point, an AI art agent used GANs (Generative Adversarial Networks) to create generative art that could be used commercially. The agent was trained on a variety of datasets, including photographs, paintings, and drawings, and leveraged latent variable optimization techniques to generate a variety of outputs. The results were impressive and provided valuable creative fodder. These questions arose both in the sense of legal rights to the generated art, but also in the sense of moral ambiguity around whether the art was a true representation of what is defined as "art" in the human sense.
- Washington Post article, "Who Owns AI-Generated Art?": https://www.washingtonpost.com/technology/2020/04/28/ai-art/ - Trevor Paglen's official website: https://trevorpaglen.com/work/image-net-roulette
- Shane Reiner-Roth, "An introduction to the ethics of AI in the art world": https://www.archpaper.com/2019/12/an-introduction-to-the-ethics-of-ai-in-the-art-world/