The year 2022 has definitely been marked by the advent of images generated by Artificial Intelligence. If at the beginning the first renderings looked more like failed experiments, now certain models and algorithms are able to generate images, from which it is sometimes impossible to tell whether they have been created by a human or not.
Some see it as the arrival of more powerful tools that will make artists even better, same as a new digital effect in a Photoshop update. Others see it as a direct threat. At the end of 2022 you could not miss the mass movement on Artstation where many artists decided to post “No To A.I.” to strongly voice their opposition against images generated by artificial intelligence on artist sharing platforms.
Apart from the domain of digital images, AIs are also capable of generating text, music and soon, why not, videos. So what about this technological revolution? Should artists really fear the development of A.Is? But how can an A.I. create art?
Some argue that artificial intelligence does not really exist and prefer to insist on the idea of deep learning. With the exponential computing power of machines, they are now able to analyze and dissect millions of images, texts and music available in our databases (from the Internet) to determine keyword associations with existing content, by analyzing the shapes, compositions and colors of the images. Currently, there is an open source base code that you can "train", i.e. you can give the existing model a specific data bank to analyze in particular. The best example is deepfake. You train the A.I. to recognize the faces of two different people and it will then be perfectly able to replace the face of one by the other in a video. A method increasingly used in Hollywood to rejuvenate the face of actors for example.
As for the creation of images, you can therefore use an interface in which you will type a prompt where you can describe as precisely as possible what you would like to see in your image and the A.I. will take care of drawing from the vastness of the available data to propose iterations of visuals corresponding as closely as possible, to your description. And it's really sometimes very surprising, because ultimately the images offered by the A.I. are created from scratch to match the request.
Here is an example I generated on Midjourney with this prompt:
the portrait of a cat, wearing a ruff around his neck in the style of a dutch painting in the 17th century. The cat is an aristocrat.
Since the image is generated by the machine and not by a human, can we then speak of Art?
This year was marked by the first controversy around a work generated by an A.I. to have won first prize in an Art competition. This is”Théâtre d’Opéra Spatial” generated by Jason Allen via Midjourney.
He won the blue ribbon, the first prize in the digital category of the Colorado state fair in the United States. There was a huge backlash online, some said it’s the death of art and demanded that the prize be withdrawn from him because he would have cheated. Except that when the work was submitted, the jury knew perfectly well that it was generated by an A.I. and confirmed that they knowingly wanted to award it the first prize. During an interview, Jason Allen is quoted as saying, "It's not going to stop, art is dead. A.I. won, humans lost." by specifying that a ethical issues should not be blamed on a new technology. Ethics is in the hands of humans and it is the use that will be made of it that should be scrutinized.
But then, is the simple fact of being the author of a prompt to an A.I. enough to make anyone an artist worthy of the greatest masters?
You have to ask yourself what an artist is. An artist is a person who feels a desire, a need to communicate an idea, a vision and emotions using aesthetic proposals that will be either experimental or anchored in more academic creative codes. By academic I mean mastering a style, knowing how to use knowledge and techniques accumulated over centuries, such as anatomy to reproduce perfectly proportioned figures, not to mention physical techniques, such as the proper use of an oil painting, mixtures in watercolor and mastery of knives to carve wood or marble into sculpture. Figures of speech in literature and the innumerable processes of composition in music. All the knowledge and collective memory of our cultures through the history of art allow us to define on a daily basis what art is, its codes and how to learn it. All of this knowledge can certainly be learned by machines. But the supreme Grail of the accomplished artist is to find a way to emancipate oneself from it experimentally and to transcend all previous forms to maintain the perpetual movement of creation. We may think of the revolutions of Impressionism and Cubism, for example, to name only the best-known movements in our recent history.
So does the A.I. feel this need to communicate an emotion? The A.I. is not yet able to spontaneously generate its own content with the personal ambition of wanting to offer its most intimate emotions to humanity. Ultimately, it still needs a prompt, a request thought out by a human. And can it really innovate and create new art forms or will it always be limited to the man-made ones available in its database? Is it capable of transcendence and emancipation? Is it even capable of abstraction?
So would the problem be solved if we assume that an A.I. cannot be an artist by definition? Since it is not endowed with its free will and is content to produce what humans ask of it, ultimately a bit like a super search engine?
But there is the problem of intellectual property. Indeed, when the A.I. is trained to reproduce images, it also analyzes and stores images of artists, dead or still alive, faces and photographs of existing people, promotional images belonging to brands or music belonging to record companies. For example, if you ask the A.I. to produce an image in the style of Monet, it will be able to create an original image, which Monet himself could have painted. But then, is it the intellectual property of the original artist? Do I have the right to use this image claiming that it belongs to me? Or is it the intellectual property of the authors of the algorithm because it is thanks to their code that the image could be generated? Or even of the person who writes the prompt? Is it valid if you explicitly cite the name of another artist in your query?
All of these questions are fascinating. Recently, after the death of Korean artist Kim Jung Gi, someone trained an A.I. exclusively with this artist's work and it is now able to generate new images that look just like the original works of Kim Jung Gi as if he had been resurrected. An approach that was very badly received by the public.
And it is behind these copyright considerations that the majority of artists on Artstation have decided to stand up against the A.I. by posting “No to A.I.”
To conclude, I understand the growing concern about these technologies. But if you're a true artist at heart, I don't think that becomes a problem. On the contrary, the A.I. can even become a tool to produce faster. Just imagine describing what goes through your mind and seeing it come to fruition before your eyes in seconds? What if you did a very specific curation of your own works, and your inspirations that only you know how to dig up to train your A.I. in a very specific aesthetic and ensure that you continue to produce unique visuals? You will remain the person in charge, who corrects, selects and approves or not the proposed visuals to ensure that you can offer your vision as the only end. In any case, that would be an optimistic theory.
In my opinion, those who are most afraid of the A.I. are the ones who know deep down that they are no better. I am thinking of all those who reproduce styles and genres seen and seen again by following trends, those who make fan art without having their own universe, those who even before the arrival of the A.I. seek their inspiration on Pinterest and in the work of others, always on the edge between plagiarism and inspiration. Those who respond to commissions as one would send a prompt to the A.I. precisely by imitating styles defined by customers. Many talented artists are caught in this game because it represents a significant source of income. For once the A.I. will definitely reshuffle the cards and pull the rug out from under those who have nothing better to offer.
Only artists who will really have something new to show will be able to break through, but as it always has been the case in each era. Another category of people who need to worry about A.I. are graphic designers, digital technicians, those who perform executive tasks under the orders of other artists or project managers. I am also thinking of all the concept artists on film or video game productions. The only thing that can save us is to develop deep human skills and probably we will see a resurgence of the attachment to physical works as opposed to digital.
So deep down, yes, I'm scared. But I'm not afraid of the A.I. I'm just scared because the path to becoming a real artist has definitely become more and more tenuous. I'm afraid like Baudelaire was afraid for painters when photography was invented. I'm scared because I know it's deeply harder than you think to really be an artist and realize that A.I. will redefine a large part of us to the rank of craftsmen with whom it will enter into unfair competition.
So it's time, more than ever, to develop your creativity, your style and seek inspiration far from the machines.