Can artificial intelligence really create art?

Can artificial intelligence really create art?

Klaus Mitteldorf, Brazilian artist of German and Swedish descent, internationally recognized for his transdisciplinary work, helps propose an answer

As a researcher on the subject of artificial intelligence (AI) and given its advances, I ended up participating in some debates around the artistic capacity of generative AIs such as Dall.E 2 and Midjourney. Induced by the quality of the images generated by these tools, some people argue that AI can actually create art.

I have no doubt that AI has the ability to generate images, poems or music that make us conclude that they are works of art, even knowing how such algorithms work: based on a large volume of human creations, they learn a pattern of which is accepted in texts, music or images, and, based on these templates, generate variations. Therefore, in a way, they learn the characteristics that the result needs to have to be accepted by human standards, and thus convince us of its legitimacy. In other words, such algorithms learn to “deceive” us, in a way. However, regardless of the impressive results, would this really be art?

To help me answer this question, there is nothing better than talking to those who actually produce art. This led me to speak with Klaus Mitteldorf, a Brazilian artist of German and Swedish descent, internationally recognized for his transdisciplinary work, including photography, cinema, collaborations with musicians, and conceptual artistic experiments seeking to expand the boundaries of these forms of artistic expression. His expression covers a wide variety of themes and styles, ranging from sophisticated fashion to urban culture and social activism, such as the album covers of the rap group Racionais MC, where he used photography to capture the essence and message of the members. of the group.

For Klaus, art is something essentially human, as it is the materialization of the expression of the artist's feelings. He sees art as a universal refuge for human beings, a means by which artists convey their deepest inspirations and sensations.

When asked whether AI can create works of art, Klaus is skeptical. He recognizes that AI can generate works based on existing data and trends, but questions its ability to evolve, taking the next steps on its own, just as artists do, as they do not have access to the deepest feelings, not being able to, in his words, “define a person’s present moment, much less their future”. To reinforce the superficiality of AI, he adds that “we have secrets stored that not even we ourselves can access, much less AI.”

According to him, therefore, AI lacks essential elements for genuine art, being just another tool, among several others, in an artist's repertoire. As a tool, it does not produce art, but rather it holds the creative and emotional essence that are the driving forces of artistic expression, and is therefore exclusive to the artist. In his view, AI can be a means to develop new technologies and expand the reach of art, but it should not be considered a substitute for human artistic creation.

After learning his perspective, I decided to try to get AI's own perspective on the topic. To do this, I uploaded the transcript of the conversation I had with Klaus on ChatGPT 4.0, and asked “Do you think AI can generate art? How does your view on the topic compare to Klaus’?” I got the following response:

The question of whether AI can generate art is complex and multifaceted. From a technical point of view, AI has already demonstrated the ability to create works that can be classified as art, using algorithms to generate images, music, texts and other forms of creative expression. These creations are often based on large volumes of data and patterns learned from existing works of art, allowing AI to produce works that, at least superficially, resemble human art.

However, the question becomes deeper and more philosophical when we consider what constitutes “art” in its truest sense. Art is often seen as an expression of human experience, emotions, thoughts and worldviews. It is intrinsically linked to the artist's ability to transmit something of himself – his emotions, his experiences, his cultural context, his worldview – through his work. In this sense, many argue that AI, which operates on data and algorithms without life experience or emotion, cannot create art in the same sense as humans.

Klaus Mitteldorf's vision, as expressed in the interview, resonates with this idea. He sees art as a deeply human expression, linked to the artist's ability to communicate their unique experience and emotions. He appears skeptical about AI's ability to achieve this level of expression, as, in his view, AI lacks the “soul” and life experience that are central to human artistic creation.

Upon getting the answer, I asked “Do you think it is important that AI one day can make art in the deepest sense?”, and I was surprised by the answer I got:

The question of the importance of AI one day being able to make art in the deepest sense is intrinsically linked to the way we value art and the role we attribute to technology in our lives.

On the one hand, if AI were able to create art that deeply expressed experiences and emotions, it could broaden our understanding of creativity and the nature of art itself. AI that achieves this level of creation could offer new perspectives and inspiration, potentially collaborating with human artists in innovative ways. This could also lead to a better understanding of AI's cognitive and emotional capabilities, opening new frontiers in areas such as emotional artificial intelligence and machine learning.

On the other hand, art is often valued for its connection to the human experience – it is a way of expressing what it means to be human, with all the complexities, emotions and perspectives that entails. If AI were able to replicate or even surpass human art in depth and expression, it could raise philosophical and ethical questions about the uniqueness of the human experience and the relationship between humans and machines. There is also a risk of devaluing the human element in art, where authenticity and personal connection can be overshadowed by technological productions.

Ultimately, whether or not it is important for AI to achieve this level of artistic creation depends on how we value the interaction between humanity and technology, and what we consider to be the essential core of the artistic experience. For many, art will always be seen as a uniquely human expression, while others may embrace the idea of ​​a creative collaboration between humans and AI. The evolution of AI in artistic creation will continue to be a fascinating and multifaceted topic of debate.

As noted by Klaus, it is essential that we distinguish repetition from innovation. Although AI can replicate structures, it still lacks the spark of authentic inspiration that often distinguishes masterpieces. In other words, AI only generates the tangible component of a work of art: what we can observe. However, we cannot consider his creations as real artistic expressions because they do not contain an essential component of art: motivations linked to human emotions.

As a researcher in the field of AI, I believe that science will develop increasingly sophisticated AIs, with the ability to even emulate motivational aspirations based on synthetic emotions, and which, like deep fakes, will deceive us very well. However, he believes that such creations will never replace the essential human component for them to become true works of art.

However, this does not invalidate the role of AI in the artistic universe, as its evolution is allowing the creation of new tools that democratize and enrich the capacity for creative expression. Generative AIs currently allow, for example, anyone, even without the talent of an illustrator, to generate images by providing only textual descriptions of what they want or imagine. Thus, in the same way that the brush and the canvas do not create art without the artist, AI does not generate the illustration without human inspiration described in a textual representation of what it wants. And in this case, it seems to me that we can say that the seed would be the text originating from human motivations, most likely linked to emotions. As, in this case, the human component is present, could we consider the result an artistic expression?

AI will not replace humans in art. However, the synergy between humans and this new tool will allow more people to take the risk of expressing their emotions in formats they would never otherwise risk due to lack of training or skill. As such tools continue to evolve, this democratization could lead us to new expressive styles that are still unimaginable, where technical innovation and emotion merge. I believe we will be positively surprised by this partnership in the future.