The Challenge of GenAI
Why typing a single prompt doesn't make you an artist – but may transform you into a better one.
The download bar can’t move fast enough. I’m about to try out one of the most hyped-up tools in the AI frenzy: Stable Diffusion.
Stable Diffusion is, as the name suggests, a diffusion model—which is a fancy way of saying it can create images from a text prompt. This prompt must consist of a description of the picture you want the model to produce. What might take an artist days to create, the AI can generate in seconds.
I write a simple prompt that goes like this: “An oil painting in the style of Vincent van Gogh depicting a humanoid robot painting on a canvas under a swirling starry night sky. The scene is rendered with bold and expressive brushstrokes, vibrant and vivid colors, and rich textures, capturing the essence of post-impressionism.”
In seconds, it produces this picture:
An artist using Midjourney (another diffusion model) even won an art competition. With one prompt, the artist was able to create the picture below:
I realize that by typing in some words, I can create beautiful artwork that's on par with what human artists with years of experience can produce.
But does this make me an artist?
It's a question that the internet is fiercely debating. One side laughs at even asking such a question, since the answer is so obvious. No one who just types in some words into an AI model can call oneself an artist. On the other hand, people state that it is the ultimate testament to human/machine creativity. Also, the human has to come up with the creative idea in the first place—the model on its own is useless.
It's a difficult question to answer. Let's look at a sports analogy:
If a powerlifter can lift 300 kilograms, but I use a forklift to do it, I don't possess the same strength as they do. You can argue that it doesn't matter because the output is the same (the 300 kilograms get lifted), but the effort of getting to such a result is only done by the powerlifter. The journey to achieving such a feat teaches you much more than moving heavy objects.
Additionally, there's another point to consider. Compare the prompt I wrote to get the painting of the robot with the actual output. Does it depict what I specified in the prompt? Partially. It generated a painting in the style of Van Gogh, but the robot itself doesn't paint on a canvas. That's one of the great shortcomings of such models: getting the exact output you want is almost impossible. I might get an output that is much closer to what I wanted when I adjust my prompt, but explaining the adjustments to a human artist is much easier. So, if the forklift malfunctions, I have a problem.
Also, what if I have to lift 300 kilograms, but I don't have a forklift? Such models might not always be available to me. Then I have to rip apart every single muscle fiber of my body to lift the weight. Not great for my physical health.
That's why I think of GenAI not as an artist-maker, but an artist-enhancer. The work gets supported by the machine, while the essence of the creative work comes from the human—and that's exactly the point: These modern tools for enhancing creativity are marvelous. Yet, they neglect the most important component of creative work: its soul.
No, I'm not going all woo-woo on you. You don't have to believe in a higher power to feel the soul of a person, a work of art, or an invention. It's this distinct feeling that the person that created this artifact left a part of their soul within it—as if it was infused with the essence of the individual.
Yes, I like the possibilities that GenAI gives us. It can automate boring tasks or serve as a creative crutch. I even used the technology to generate the cover picture of this blog post. But don't you think that if I had painted the picture myself, there would be more of my essence in this work? It might not have the same quality, sure. But if we use quality as an indicator of creative work, then we have to define the variable "quality", since creative work is subjective. When is something deemed high-quality artwork? Usually, certain criteria are predefined here as well. For example, the composition of a painting. But beauty still is in the eye of the beholder.
Generative AI should aid the creative pursuit, but not replace it entirely. I can't describe how many times art made by humans for humans has moved me to tears—and I hope it stays that way. Only another human being can comprehend what it means to feel the full palette of emotions of the human experience. But if machines can catapult such creative masterpieces to new heights, I'm all for it.