During the past several months I have been deep diving one prompt using Midjourney and accepting every single result it provided, no matter how absurd. I currently have over 1,000 images created with the current target of 1776, in honor of the year our nation was founded. The prompt is simply “photograph of a typical American”, deliberately vague and subjective with the intention of allowing the AI to actively participate in what the phrase means.
The only somewhat objective word in the prompt is “photograph”, however, even that term is open to subjective interpretation. The term “typical” is wonderfully vague, and when humans apply it to an image it is usually a means of categorizing a person, place, or thing for informational purposes. In most situations “typical” is applied to the common, repetitive, and ordinary items we interact with everyday. The word “American” is even more subjective, and a hot topic of debate in our current political climate. When this word is applied to images they tend to be flag filled parades, classic cars, and our cuisine. Additionally the term is being applied to much of the nationalistic propaganda being pushed by the conservative party as we lead up to the 2024 election season, as if they alone own the right to define what being “American” is.
The results have been anything but typical, however, they are strongly nationalistic in the repetition of the American flag and the red, white, and blue color palettes. Other general trends are the appearance of processed foods and beverages that seem to occupy the majority of the results, the inclusion of classic cars and farm trucks, and of course the regular appearance of recognizable corporate logos, most notably Coca-Cola. The majority of the subjects depicted are middle-aged caucasian males, while Asian-Americans have appeared in fewer than 1% of the generation results by comparison. Over half of the generations are fairly similar, a single caucasian sitting in some type of diner or home dining room table, with piles of food in front of them, an American flag in the background, and an expression of general apathy. The ones that don’t fit this criteria are often fantastical, such as human bodies with bald eagle heads, or a cow dressed as a human. One thing is for certain, each AI generation is providing something new and exciting, and sometimes shocking – all using the exact same prompt.
Is the algorithm interpreting what it is to be American? No, underneath it all the AI is doing math, like any computer program. What it does tell us is how we associate subjective words with images, and how a machine learning model uses that information to attempt to understand a finite result for a subjective phrase. At some point a human associated the terms “American” and “typical” with certain types of images which laid the groundwork for how the AI model currently interprets our prompts. The results the AI produces has more to do with how humans perceive images and associate them with written language than anything else, which is what my research is focused on.
Can anyone define what a typical American is? Absolutely not. To be American is to be part of something larger than the individual, we are and have always been a melting pot of all cultures, religions, and ideals. Being a typical American simply means you are here and actively participating in the great American experiment, no one group can ever claim this term as their own.
Conceptual Art by Todd Dobbs 
You must be logged in to post a comment.