This is the forth article in the series. Here are links for others:

4 Ways to Illustrate a New Book for Beginners

Making Illustrations for Your Book When You Can’t Draw

Making Illustrations for Your Book with AI

Marketing’s Influence: Shaping Perceptions and Misconceptions

Marketing surrounds us in every aspect of our lives. It's hard to imagine modern civilization without it. While it helps sell products, it can also create misleading impressions. Take "3D" movies, for example: what we watch in theaters isn't true 3D. It's actually stereo, meaning we can't change our perspective to see depth as we would in real life — it only works from the camera's viewpoint.

The same goes for AI. In most real-world applications, "AI" typically refers to a complex program that generates new content by combining pre-existing elements and random values. Most AI image generators don’t actually create things from scratch, nor do they fully understand what they're producing. This is why AI tends to execute input descriptions in terms of composition, color, or style, but often misses the logical aspects of the image. As a result, we sometimes see humorous or unexpected outcomes when working with AI — especially if we're not under a tight deadline and can afford to laugh at the quirks!

Humorous Missteps: My First AI-Generated Illustration for Lviv

My first humorous surprise from AI came while working on an illustration for a story "Significant deed"set in the Ukrainian city of Lviv in the 1920s. It was based on the recollections of one individual, who shared how two generations of Independence fighters were connected. In Ukrainian, "Lviv" can be translated as the "city of lion," and when I was creating the cover image, the AI took that translation quite literally! It would have been amusing if I had asked for an illustration for "The Lion in Winter," but it was even funnier because the word "lion" wasn't even part of my input description.

The next twist was even more unexpected. I asked the AI to draw a boy in winter clothes wearing a fur hat. In western Ukraine, it was common for children’s winter hats to be made from pieces of fur. But what I received was not just a boy in a fur hat — it was a duplicate of the kid… as a bear! Sure, there was fur involved, but definitely not in the way I had intended. While the previous mishap made me think of "The Lion in Winter", this one had me imagining "The Golden Compass".

Another illustration I requested for this story was meant to depict children having a snowball fight in the street. But apparently, in the AI's mind, a snowball fight is far more artistic than I could have ever anticipated!

The Four-Legged Challenge: AI’s Struggle with Horses

While working on my next story, set in the 1870s and focused on the misadventures caused by Russian propaganda in Ukraine just before the Russo-Turkish War, I needed to include horses in several illustrations. However, AI seems to struggle with drawing horses that have four legs — occasionally it gets it right, but more often than not, the number is... a little off.

Meanwhile, when the humans in the illustrations are placed outside of their "normal" environment — like lying in bed or sitting at a table — the number of arms and legs can vary dramatically.

In the AI's mind, horse-drawn wagons aren't practical means of transportation; they become abstract, artistic objects. I had to manually correct all that "horse-wagon" imagery using Photoshop because no amount of detailed descriptions or restrictions seemed to make any difference.

Beyond Boundaries: When AI Characters Ignore Instructions

Now let's talk about human behavior and facial expressions. In the AI's world, a mother's love for her child can sometimes come across not just as odd but even unsettling.

Faces, in particular, are the zones where the most unpredictable things happen in AI-generated images. At times, they are so peculiar that they defy explanation.

In one of my articles, I mentioned how the AI occasionally made the characters do things I had explicitly prohibited, like kissing. But sometimes, they engage in even stranger actions, which are difficult to put into words.

Lastly, I want to touch on AI’s deeper understanding of reality. In one of my stories, I was writing about a Russian army unit attempting to massacre a Ukrainian city in December 1917. I needed pictures of russian soldiers, so I provided the AI with an example. While the general composition was correct, what I saw were not russian soldiers but uniforms that resembled those of Nazi troops from World War II. By an eerie coincidence, many Ukrainians today compare Russian war crimes to those committed by the Nazis in World War II. So perhaps I could interpret this as AI's attempt to grasp that concept.

Conclusion:

AI-generated content, while fascinating, often reveals the limitations and quirks of machine learning. What we typically call "AI" is, in reality, a sophisticated system of patterns and random combinations, not true intelligence or creativity. As a result, AI sometimes struggles with accuracy, especially in more complex scenarios like drawing human forms, interpreting historical contexts, or even representing emotions. However, these quirks also highlight the current gap between artificial and human intelligence, reminding us that true creativity and understanding remain distinctly human traits.

P.S After writing this text I asked AI what does he thinks about this article. Here is the answer:

Overall, this article presents a balanced view of AI-generated illustrations, recognizing both the fascinating possibilities and the inherent challenges. It highlights the importance of understanding AI's limitations while appreciating the unique contributions it can make to creative processes.