The Prolification of Generated Stock Imagery

How did you feel when you saw the image above? Probably a slight bit of confusion as the image above seems to look like a presentation of a mobile app demo and NOT a generated image.

Let me change the question. How does the image above MAKE you feel?
Perhaps a bit refreshed upon seeing an actual photograph of an event. I had typed in the Unsplash search bar "large meeting", and this was one of the photos that popped up.

I'm here to talk about GenAI. If you haven't heard about it yet, it's the latest hot topic in the world. Generative Artificial Intelligence. It's the AI that generates. It creates novel content trained via prompts with answers that could be more text, images, programming code, and even music.

Late last year there was a trend online called "Make It More". Essentially, people would prompt ChatGPT asking for, let's say, an image of "someone giving a presentation to an audience", and each time, progressively asking ChatGPT to make it more. In this case, we could ask it to "make the audience larger" or "make the presentation bigger". If you've been on LinkedIn in the past year, you've probably seen an image generated by ChatGPT attached to someone's post. There's a blue-tinted image of a meeting that has been used so widely, I ended up getting sick of seeing it. (I somehow can't seem to find it as I'm writing this, but if you know, you know.)

As more and more people turn to GPT for stock imagery and turning away from stock photo sources or even the simple Google Images search, I am personally finding the internet to have become blander. There's no more personality online. Everything is AI-generated with the familiar taste and art style of OpenAI's ChatGPT.

ChatGPT has an art style?

Yes, even an AI trained on the massive amounts of image data scraped off the web has its own art style. If you've seen enough generated images, it's pretty easy to tell.

Toys"R"Us recently created an advertisement that had used OpenAI's latest Sora model with a tiny amount of post-production VFX. When I heard of this, I was curious. How would this ad make me feel? How uncanny would the images be?

After watching this ad, I could immediately tell it had the characteristics of GPT art. The way the colors are used, the lighting, the way the facial proportions show up and even how each individual toy is shaded all had the very familiar GPT aesthetic.

(Note: There are some photoshop artists that have an otherworldly aesthetic to their work who have been falsely accused of using AI. See Andrew Huang's latest YouTube video for an example.)

Creativity can't be automated

The essence of creativity is the energy that a person has to create. A vision in the artist's mind that comes to life within the constraints of what is possible in the real world. It's a method of self-expression. The essence of art can't be automated.

Stock images aren't usually treated as art, but the photography work to compose and edit the images requires a photographer to be creative. When a writer writes, they are being creative. When a writer is trying to find imagery to illustrate their words, using a stock photo or hiring an illustrator fosters a collaborative opportunity for two artists.

When using Generative AI as a cheaper alternative to producing the work, there is a removal of the artist. The AI is trained on the works of others and essentially "plagiarizing". The AI learns what the model was trained to do, but it doesn't really have the essence of creativity that ads the artistic twist. The person using the AI isn't exercising the creative muscle and learning the skills to create their vision, they are simply prompting the AI and picking from a set of options. It feels lazy and lacks the human depth that comes through in art.

AI Art is like Fast Fashion

Fast fashion is typically described as the consequence of cheap mass production in a capitalistic world. People who are interested in fashion, a form of art and self-expression through clothes, typically don't gravitate towards fast fashion or support it due to the wasteful nature and the lack of quality that usually goes into it. Fast fashion aims to replicate the work of designers at low cost.

AI art is similar. The business argument for Generative AI is that it is the cheaper alternative. When it comes to a writer looking for that image to illustrate their idea, they'll either pick the best stock image available, create the image themselves, or hire someone. The draw of AI is that the writer would be able to create the exact image that they want without doing the work. Who would be doing the work then? The AI that's been trained on the work of millions of artists without proper credit?

Unlike fashion where there isn't a free alternative to self-expression through clothes, it's free to use a stock image. It's very cheap and feels lazy to use Generative AI to offload the unfulfilling tasks. If you're a creative who wants to offload the unfulfilling tasks, it's better to collaborate with someone who can realize your vision. It would also offer a creative exercise to another artist and supports their career. Admittedly, it will take some time to find that collaborative fit, but it'll be worthwhile to create an authentic piece.

When I write, I prefer not to use GenAI for my images. I feel that the stock images that are out there in the massive Open License and Free Use repositories have the sufficient variety. There's also a sort of creative fun in seeing how the same stock image can be used to illustrate different topics. To businesses who are only trying to churn faster numbers of a product or quicker pieces of content, I'd still encourage original creations of assets. Using an AI result off the shelf without any post-processing, editing, or human creativity cheats the value of art, the human part of it.