Рубрики
Статьи

Why Artificial Intelligence is not a problem

For quite some time now, I’ve been fascinated by neural network-based artificial intelligence (AI) and have been using it in the process of creating my illustrations. I’ve already described one of the ways I use AI in my first article dedicated to this topic. Now, I’d like to expand on this subject and talk about the broader applications of AI in an artist’s work. I’ll also share some thoughts on issues that spark the most heated debates in the art community.

AI-assisted artworks

Some of my illustrations can be categorized as AI-assisted to varying degrees. This means I incorporate individual elements generated by neural networks into my work. I find this format most suitable for me for several reasons:

  • I can’t fully explain my ideas to the AI using text prompts alone. I’m not a writer or a programmer, so entirely AI-generated images don’t satisfy me. They don’t align with the vision I have in my mind.
  • I don’t feel the need to manually draw every element in my artwork. I’m not someone who creates art purely for the joy of the drawing process itself. As a result, works created without using AI-generated elements often leave me less satisfied.

Many of my illustrations are surreal in nature. Achieving this style is most effective for me through a synthesis of my own ideas and machine-generated elements.

That said, I do have works without AI-generated elements that I find satisfying. Sometimes, my characters or compositions are too abstract or specific for the neural network to understand, and there can be nothing from the generated outputs that I can use to improve them. However, working without AI also offers a valuable opportunity to view my work from a fresh perspective (as simple mirroring doesn’t help me).

We’re All Going to Be Replaced!

As I mentioned in my first articleAI finishes your work instead of you only if you don’t know what you want. If you do know, it will still finish it — but not in the way you want.

For my AI-assisted artworks, I start by creating a fairly detailed sketch. I need to personally decide on the format, arrange the objects within it, build the composition, and select the color palette based on how I imagine the final image. My artistic knowledge also comes into play when evaluating the generated outputs, as I need to determine how closely they align with my vision and which (if any) elements from the AI-generated images I can incorporate into my work. (As I’ve mentioned before, sometimes none of the generated elements are usable.) I also rely on my skills and understanding to integrate the AI-generated elements into the overall composition, draw the parts I can or want to create myself, and finalize the piece so it looks cohesive.

An artist will remain irreplaceable as long as they maintain their value. It’s crucial for us as artists to understand where our value lies so we don’t lose it:

  • Physical experience: Knowing how objects look and function in reality.
  • Life experience: The ability to analyze it from different perspectives.
  • Formal knowledge: Understanding the principles of creating images and how they work.

To achieve high-quality results, a neural network operator ideally needs to know how to draw themselves — or, at the very least, possess strong image-editing skills, an understanding of composition and color theory, as well as a well-trained visual library and good observational skills. They should have a clear understanding of how objects appear in reality, how to stylize them effectively, and which visual techniques provoke certain viewer reactions. They must also be able to critically evaluate AI-generated images, rather than treating them as infallible marvels.

And this brings us to the core issue of modern (de)generative art.

Low entry threshold

The average internet user typically doesn’t possess artistic skills or knowledge — and that’s perfectly fine. For most people, these skills aren’t necessary for a successful life. However, their perception of creativity and art often remains at a primitive or clichéd level. When such a user gets their hands on a high-level, almost magical tool, they lack the ability to critically evaluate its results. Their judgment is usually based on simple criteria like: “I couldn’t draw this myself!” or “It’s bright, cute, and detailed, so it must be good!”

This is precisely why many AI-generated images that get shared online cause rejection among us, artists. The issue isn’t that these images are created by a neural network, but that they are often visually low-quality works — lacking depth, aesthetic harmony, and meaning.

At present, in the public consciousness, neural networks occupy a space between a toy and a magic wand. The task for artists is to take control of this tool and demonstrate that it’s not an all-powerful magical artifact, but rather a specific professional tool designed to solve particular types of tasks. Moreover, it’s up to artists to preserve and uphold the concepts of aesthetic and visual harmony. We need to actively influence the development of public taste to ensure that the trajectory of visual culture remains at a high level.

Artists should also unite to address the moral and ethical questions surrounding generative AI.

Legality and Ethics

The legality of the existence and use of generative neural networks is a topic that raises many questions and often indignation.

Can you imagine artists of the past using neural networks? Of course, back then, they were called something else — “apprentices” or “assistants”. Yes, I’m talking about biological neural networks. For example, contrary to popular belief, Michelangelo didn’t lock himself away in the Sistine Chapel so that no one could disturb him while he worked on the ceiling frescoes. This wasn’t possible, not only because the chapel had to be used for divine services, but also because it would have been unrealistic for one person to complete such a monumental task within the given timeframe. Especially since Michelangelo considered himself primarily a sculptor and had not practiced fresco painting for many years at the time he took on the commission.

As a result, elements such as the architectural imitations, backgrounds, and cherubic putti on the chapel’s ceiling were not painted personally by Michelangelo. Yet, they still became an integral part of the iconic artwork.

A ceiling of the Sistine Chapel

Take, for instance, Ivan Shishkin’s famous painting «Morning in a Pine Forest». The bears in this painting weren’t painted by Shishkin himself but by another artist, Konstantin Savitsky. What’s more, Savitsky wasn’t an assistant but a full-fledged co-author of the idea. This raises an interesting question: Did Savitsky “generate” the bears for Shishkin’s painting, or did Shishkin “generate” the forest background for Savitsky’s bears?

«Morning in a Pine Forest» (or initially «Bear Family in a Forest»)

The image of the lone artist, creating masterpieces entirely on their own, is actually a relatively modern concept. It emerged only in the late 19th century, with the advent of Impressionism. Impressionist paintings, being representations of a single person’s impression (impression) — the artist’s own — did not involve contributions from others. However, artists of the Middle Ages and the Renaissance often worked in large workshops, and completed works were not always signed by the individual who created them. This was partly because multiple people could work on a single painting.

For example, before Caravaggio began his independent artistic career, he was employed to paint flowers, leaves, and fruit, contributing details to paintings mass-produced by Cesare d’Arpino’s workshop.

Skeptics might argue, “An assistant consents to the use of the elements they create.” That’s true. But is consent always given in other art forms? For example, does an illustrator explicitly consent to their work being used in collages? Any person who buys a book, magazine, or brochure can cut out images and create their own art from them. This is exactly what the artist Max Ernst did in his visual novel «A Week of Kindness». Ernst not only failed to credit the authors of the original illustrations but to this day, many of them remain unidentified by art historians.

A fragment of a visual novel «A Week of Kindness»

Today’s digital artists have access to a variety of tools, but are these tools always obtained legally? Alongside Microsoft Office, Adobe Photoshop ranks as one of the most frequently pirated software programs. Additionally, stolen brush packs from private Patreon accounts, textures, and stock images often make their way onto the internet, where they’re subsequently used in photobashing and other techniques. This raises questions about the legality of many artworks created without neural networks.

Rather than fighting against neural networks and their users, artists concerned about copyright issues should focus their efforts on finding solutions to protect their work. And I don’t mean memes or emotional outbursts on social media — I mean real action. Artists aren’t lawyers, but they understand the nature of the problem better than anyone. Without their resonant voices, no lawyer or policymaker will take the issue seriously.

One possible solution lies in blockchain-based technologies like NFTs. While initially created to protect copyright, NFTs quickly became a tool for scammers and lost much of their credibility. However, their potential as a content protection mechanism remains largely underestimated. Another promising option is algorithms like Nightshade and Glaze, which modify images to make them unrecognizable (or even harmful) for neural networks during training. Despite the promise of these technologies, the aggressive rhetoric of many of their proponents is concerning. Treating these tools as «weapons» in a war against AI risks alienating potential users who could otherwise benefit from them.

Artists (and really, all internet users) should also remember: everything that ends up on the internet stays there. If your image can be downloaded, copied to a clipboard, or screenshotted, it can be used to train neural networks. Furthermore, any image shared online inevitably becomes part of the cultural context. It influences global culture and shapes how people perceive the world and creativity. It remains in the consciousness of those who have seen it. You may never know this, but years later, someone might create something strikingly similar to your work after seeing it just once. Why? Because they care about the same themes you do, and your work may have unconsciously influenced their perception of that topic. Of course, their version will be filtered through their own life experience. But if they care about the same themes you do, their life experience might resemble yours, making it difficult to define the acceptable degree of “similarity.” Culture is a process of exchanging ideas, not rigidly guarding them.

I believe that, through the combined efforts of artists and programmers, we can create ethical generative neural networks. But this will undoubtedly require significant changes to the way images are published across the internet. For now, a potential solution looks like this to me:

  • Improving and increasing the accessibility of algorithms like Nightshade and Glaze, possibly even integrating them into social media platforms and website-building tools.
  • Providing artists with a legally established way to sell their images for neural network training. This would also reduce the number of free neural networks, effectively raising the barrier to entry that I mentioned earlier.

The Art of Deception

When discussing ethics, it’s impossible not to touch upon deception and the act of misleading the viewer — something visual art has done throughout its entire history. After all, an image of an object is not the object itself. This concept was famously illustrated by René Magritte in his painting “This is Not a Pipe”.

«This is not a pipe» — this is an image of a pipe.

The quality of a work of art is often judged by how convincing its illusion is. This doesn’t necessarily mean creating a realistic depiction of objects. It can also mean the illusion of an event that evokes specific emotions in the viewer or provokes certain thoughts, regardless of the materials or tools used by the creator — be it oil paint, a stylus, charcoal, or a neural network. If the artist fails to evoke the intended feeling in the viewer, it merely indicates a lack of skill.

There is another common form of deception, whose ethics are questionable. Is it ethical to convince viewers that every human-created image inherently contains a soul, and that its absence is what makes AI-generated works inferior? This claim, to put it mildly, is dubious. For example, commercial artists working on stock platforms often recommend uploading 3–5 illustrations per day to achieve a stable income. This turns their work into a sort of assembly line, where speed is critical to financial success.

Modern commercial illustrators must adapt to the fast pace of life and content consumption. Their workload is far more intense than that of artists in the Middle Ages or the Renaissance, who worked in large workshops. Under such conditions, there is no more soul in each individual piece than in a factory-produced glass.

When you think about it, this is a natural progression. Over time, manufacturing workshops were replaced by factories with increasing automation, as machines are perfectly suited for mass-producing standardized objects. However, despite this, glassblowing and pottery workshops still exist today, crafting unique items. These creations hold less utilitarian value but are perceived as having soul due to their artisanal nature and the personal touch of their creators.

Art will not disappear with technological advancement. But it’s important to recognize the difference: mass production of glasses is not creativity. We shouldn’t deceive ourselves or others by attributing soul to every commercial illustration. In the context of mass production, whether by humans or machines, functionality takes precedence over artistic depth. The true soul of art will remain where there is space for uniqueness and personal expression.

Conclusion

Like all things new and unfamiliar, generative neural networks provoke anxiety, fear, and, as a result, sometimes aggression born from the need to protect oneself and a familiar way of life. However, fear of the unknown can be overcome through a deeper and more thoughtful understanding of this new phenomenon.

It’s important not to succumb to emotions — whether fear, despair, or euphoria. Instead, assess the situation with a clear mind. Look at the objective advantages and disadvantages of the new technology. No technology is inherently a salvation or a curse. Its status as either depends solely on how it’s used. The more complex our technologies become, the more risks they carry. It is crucial for people to remain responsible users of their tools. This requires awareness and effort, but such is the cost of progress. Progress must be guided and controlled, not stopped. Otherwise, we risk losing even the technologies we already have.

Seeking help and using tools to make work easier is normal — not just for artists, but for people in general. The choice of assistants and tools always remains a personal decision.

While preparing materials for this article, I came across the website AIArtists.org, a community of people interested in using generative images in art and exploring the related moral and ethical questions. If this topic intrigues you, I highly recommend checking out the materials on the site.

Artificial intelligence, in and of itself, is not the problem. People make it one.

Добавить комментарий

Ваш адрес email не будет опубликован. Обязательные поля помечены *