Editor's Note: When I first started using AI-assisted art, using artist modifiers was one of the most satisfying experiences I encountered. With just a few words, an ordinary image could be transformed into something magical. However, when I learned how living artists view AI-assisted art, and why, that joy of creation was never quite the same. I made the decision then to archive my works using living artists as modifiers, and to only use non-contemporary artists as artist modifiers. Advances in the technology of creating AI-assisted art have made it possible to create quite compelling images without artist modifiers, but I still like to explore with non-contemporary artist modifiers as it creates connections and synergies that I enjoy, and compels me to continue learning more about past artists. In this post, I examine some of the issues with using artist modifiers and offer some recommendations for the AI art enthusiast.
In the burgeoning field of AI-assisted art, the ethical landscape is complex and evolving. One particular area of concern is the use of artist modifiers—descriptors that prompt AI to generate images in the style of specific artists. While these modifiers can produce striking and recognizable results, they also raise significant ethical questions.
Ethical Considerations in Using Artist Modifiers
The more you know about AI-assisted art, the more you realize just how complex many of the ethical considerations are. Still, some ethical and legal issues are fairly straight forward. We'll look at some of the main considerations here, starting with the more obvious and moving into more ambiguous areas.
Training on Copyrighted Works Without Permission
AI algorithms trained on copyrighted works without the artist’s permission can infringe on intellectual property rights. This unauthorized use can lead to legal disputes and undermine the rights of artists to control the use of their work. The ethics of data sourcing are paramount; using copyrighted materials without consent raises questions about the responsible use of artistic data in AI development.
Imitation of an Artist's Unique Style
When AI-generated art closely mimics a particular artist’s style, it can dilute the uniqueness and market value of the original artist's work. This imitation can confuse audiences and diminish the distinctiveness that artists cultivate in their careers. Artists often spend years developing their unique style. AI-generated works that replicate these styles can undermine the concept of creative ownership and the individuality of human artistry. Further, an artist’s style is often deeply personal and culturally significant. Imitating these styles through AI can trivialize their original context and meaning.
Loss of Artist Income
The proliferation of AI-generated art that mimics specific styles can impact the financial livelihood of artists, reducing demand for their original work and potentially leading to a loss of income. Imitation of artists' works and styles is certainly not new, but the ease of producing high volumes of AI art can saturate the market, making it harder for individual artists to compete and maintain their economic viability.
Ethical Ambiguities in AI-Assisted Art
Ethical and copyright issues around using artist modifiers in AI-assisted art are further complicated by a multitude of factors. Each AI model has its own version of an artist's style (if it has been trained on that artist at all). AI artists typically add other modifiers to the prompt in order to enhance or alter the effect of any artist modifiers. Often, more than one artist modifier is used, leading to a blending of styles in some cases or a completely new style in other cases. On many platforms, a final image may be the result of multiple iterations and processes, making the contribution of a single artist modifier difficult to ascertain, at best. These are only a few of the complexities that face the use of Artist modifiers in AI-assisted art. These questions and many others will likely remain unanswered for at least the immediate future.
Recommendations for Ethical Use of Artist Modifiers
So what do we do? The concerns raised above are relevant to the companies behind the AI algorithms and platforms as well as to individual artists using the platforms. Some companies have responded to the concerns by attempting to disallow the use of copyrighted artist modifiers. Hopefully, these same companies will also reimburse artists for past copyright infringement and modify their training procedures to avoid future infringements. For the individual AI artist, I offer the following recommendations for your consideration. You should make the choice(s) that seem best to you.
Avoid Using Living or Copyrighted Artist Modifiers Entirely
The most straightforward way to navigate the ethical challenges is to avoid using artist modifiers that mimic the style of living artists or those whose works are still under copyright. This respects the original creators' rights and avoids potential legal complications. Further, by not relying on artist-specific prompts, artists can encourage the development of unique styles and the emergence of new creative voices in the AI art community.
Use Generic Style Descriptions
Instead of specifying individual artists, describe broader artistic movements or styles (e.g., "Impressionist style" instead of "in the style of Monet"). This approach respects the collective contributions of artists within those movements without singling out individuals. Additionally, use detailed adjectives and conceptual prompts to guide the AI, such as “dream-like landscapes” or “vivid and abstract forms,” which can foster innovative and original output without direct imitation.
Cite and Credit Sources
If an AI-generated artwork is inspired by a particular artist’s style, clearly credit the influence. Transparency about the inspiration respects the original artist and educates viewers about the art’s lineage. Consider keeping your prompt "open" (i.e., visible to anyone who views the image).
Obtain Permission and Use Openly Licensed Works
For contemporary artists or those with copyrighted works, obtaining explicit permission to use their style can be a respectful and legally sound approach. This is especially true for using the artist's images for model training. Alternatively, leveraging artworks that are in the public domain or available under Creative Commons licenses ensures that the use of these styles aligns with the artists' or estates' intentions.
Mix and Create
For contemporary artists, use more than one artist modifier to create a new, unique style (this can be done with non-contemporary artists as well, of course!). You can also use modifiers that are not typically associated with the artist(s) in your prompt to create something unique. The use of sampling in the music industry provides an analogy to this approach. Musicians can ethically and legally create their own songs using samples and pieces of other musician's songs, as long as the song they create is substantially their own. Despite some specific guidelines, there are still gray areas and resulting lawsuits, but the general framework at least provides musicians with some working guidelines. Of course, AI can now also create music, but the ethical and copyright issues there are beyond the scope of this post.
It should be noted that if you decide to use the "Mix and Create" recommendation, you should proceed with caution. Results will vary, and even by mixing in different artists and modifiers, you may still get a final result that is clearly related to one artist. For example, I have found that using "starry night" in prompts on some platforms results in an image obviously influenced by Vincent van Gogh's "Starry Night", even if I didn't use Vincent van Gogh in my prompt and did use other artist modifiers (and even if I just wanted a night with lots of stars in the sky!). Also, some of the ethical concerns raised above still apply. Personally, I use the "Mix and Create" approach, but only with non-contemporary artists. This still leaves a gray area for artists who have died but whose works are still under copyright, but I feel comfortable using this approach as we wait to see how the ethics and copyright issues of AI-assisted art evolve.
Conclusion
In conclusion, while AI-assisted art offers exciting possibilities, the use of artist modifiers must be approached with care and respect. By considering the ethical implications and adopting responsible practices, we can ensure that this innovative tool enriches the artistic landscape without compromising the rights and integrity of individual artists. This exploration highlights the balance required in using AI responsibly and creatively. By embracing ethical guidelines, AI artists can contribute to a dynamic and respectful digital art ecosystem.
Post Script
My intention for this website is to have AI do a lot of the creative work. To that end, I provided a prompt in ChatGPT 4o related to this topic, and had ChatGPT create a blog post. I ended up heavily editing the result, as this issue is important to me. This post also just scratches the surface. The reader may find it interesting that the two final images in this post were created using the same exact prompt. The image "Reflections, for Cello" was created in ChatGPT and can best be described as a collaboration between ChatGPT and myself. I then copied the resulting prompt and entered that into Dall-e3 on NightCafe to get "On Being". The resulting image is related to the "Reflections, for Cello", but clearly also different. Part of this difference is due to the AI modifying ("enhancing") the prompt I entered, which creates even more questions about ownership and creativity in AI-assisted art. It is a brave new world we are living in, indeed.
Image Information
The images in this post were created by me. Clicking on each image will bring you to their home on NightCafe and will allow provide you with more information about each image.
コメント