By Gaia Giordani. Cover Image Refik Anadol, Unsupervised, Machine Hallucinations, MoMA, NY.
The great debate surrounding the role of AI in art has been going on since the mainstream introduction of generative AI tools, largely driven by the opinions of artists themselves. This conversation is polarised: most artists agree that “it’s the opposite of art” and a threat to the very existence of art. Conversely, some argue that AI functions like a collective unconscious and that we should thank AI for challenging us out of our assumptions.
AI-generated art is now accepted as a category on its own. This differentiation helps ensure that we do not confuse works created by machines—such as videos, poems, paintings, or music —with those produced for something by humans.
Ted Chiang, among other intellectuals, argues that the creative choices made by artists are fundamentally different from those made by artificial intelligence.
In an article in The New Yorker, the author reminds us that grammar is governed by nearly mathematical rules. While generative AI currently exhibits a degree of predictability that is embedded in its very nature, there is a potential for improvement in the future.
Refik Anadol, director at RAS, is one of the most influential new media artists in the world, known for his extraordinary ability to integrate visual art, data processing and generative artificial intelligence.
Ted Chiang poses a profound existential question for both himself and uss: “Could AI surpass humans in making art, much like calculators are better at addition and subtraction?”
To start, we rely on powerful calculators that produce accurate results, while Gen-AI hallucinations are the closest approximation to art because they deviate from predictability and accuracy.
The Limitations of AI: Choice vs. Predictability
Another interesting aspect is that AI has limited choices compared to the infinite options available to human artists, writers, painters, and creative minds. We operate in a realm where limitless possibilities can be explored, allowing for unpredictable results that foster creativity and originality. In contrast, generative AI tools operate within a confined range of options, strictly limited to their underlying models.
When a new technology arises, it is common for those expected to use it—who also make decisions about how and why to use it—to scrutinise it. Photography, for example, offers a quick way to create a portrait without relying on brushes, paint and canvas. Yet, it is considered an artistic medium because it allows for endless creative choices that a photographer can make, including the type of camera, lenses, and exposure settings.
For this consideration to be valid, we as a community of artists must agree that artistry stems from the endless choices available to an artist.
AI artist Botto has generated a world record in sales of AI generated artworks at Sotheby’s (over 300,000 dollars), establishing a new benchmark and inspiring artists and collectors all over the globe to appreciate and purchase AI art.
What Do AI Tools Think About It?
ChatGPT’s Insights: The Nature of AI-Generated Art
ChatGPT acknowledges the official definition of art, as “a form of human expression that conveys emotions through creative mediums” (in its own generated words). The key terms here are “human” and “emotions”, which highlight aspects that are inherently tied to people and not to AI.
During our conversation, ChatGPT further elaborated on the concept that “while AI-generated art raises questions about originality, creativity, and authorship, it has expanded the boundaries of what is possible in the art world, pushing forward new forms of digital expression”.
When discussing how AI can create art, it’s important to recognise that it often relies on technical processes, which raises the question of whether it can truly produce original art or merely result in plagiarism. AI can generate art through techniques like deep learning and neural networks. One common method is the use of generative adversarial networks (GANs), where two neural networks collaborate: one generates images, and the other evaluates them. This process allows AI to replicate the style of famous artists or develop entirely new styles.
Jason M. Allen has been the first person to win an art competition submitting an artwork made with AI. Other than his famous “Théâtre D’opéra Spatial”, he’s still creating and selling pieces.
The Evolving Definition of Art: AI’s Role in the Future
Claude, by Anthropic, makes a heartful apology, saying: “AI systems like myself learn patterns from existing human-created works and generate new combinations based on that training, rather than creating from lived experience or emotional depth. We don’t have subjective experiences or feelings that traditionally drive artistic expression.”
Gemini Advanced lists the arguments for (and against) AI as an artist, adding that “ultimately, whether you consider AI-generated creations to be “art” depends on your definition of art itself. It’s a fascinating and evolving field, and the boundaries are constantly being pushed.”
