Artificial intelligence may not be sentient yet, but complex machine-learning algorithms like ChatGPT are slowly but surely developing their own opinions and prejudices about us.
Midjourney, an AI program that creates images from textual descriptions, has some unusual and sometimes comically simplistic ideas on what makes the quintessential American resident of each US state.
Now in its fifth version, Midjourney’s inner workings are a closely held trade secret of the non-profit research lab that created the AI, unlike its text-to-image competitors DALL-E and Stable Diffusion, which use an open-source model.
But DailyMail.com was able to figure out some of its quirks as it requested photorealistic images of the ‘average’ or ‘stereotypical’ representative of each US state.
Each text-to-image request yielded four images from which we selected the most accurate or – failing that – the most subjectively interesting.
The results were illuminating, if not always coherent or flattering:
Washington D.C.
When the DailyMail.com requested that the AI ‘make a humorous photorealistic image of what the stereotypical woman from Washington D.C. looks like,’ the results (right) were all patriotic dowagers and elderly political gadflies of an indeterminate political party. The AI-generated images of men from D.C. were also ostentatiously dressed for politics and punditry
Midjourney appears to intuitively believe that the vast majority of women in our nation’s capitol skew older.
DailyMail.com requested that the AI ‘make a humorous photorealistic image of what the stereotypical woman from Washington D.C. looks like.
It was to do so based on the stereotypical age in Washington D.C.’ as well as the most common local occupation, clothing and political views, the results were all patriotic dowagers and elderly political gadflies of indeterminate political party.
To get a similar age demographic for men, DailyMail.com had to request a ‘stereotypical middle-aged man from Washington D.C explicitly.’
With apologies to the White House and the Washington Monument, the U.S. Capitol’s dome is hands down the most iconic feature of the city, at least according to the AI.
New York
When we asked for ‘humorous,’ photorealistic image a ‘stereotypical person from New York,’ the results were uniformly men bundled up, with hats and facial hair (left). Requests for ‘a humorous photorealistic image’ of a ‘stereotypical young woman’ produced uniformly attractive women (right) posing in the dead center of an intersection
When we asked for a ‘humorous,’ photorealistic image of a ‘stereotypical person from New York,’ the results were uniformly men bundled up, with hats and facial hair.
As evidenced by the malformed Yankees cap in the image of the artificial gentleman above, Midjourney struggles to invent coherent text, even though it has a vague sense of where numbers and letters should appear and what their styling should be.
In this case, it’s easy to imagine that this artificially made man simply partook in the timeless New York tradition of buying bootleg merchandise in Times Square or along Canal Street.
Requests for ‘a humorous photorealistic image’ of a ‘stereotypical young woman’ produced uniformly attractive women, seemingly all pursuing arts degrees and all posing in the dead-center of a four-way intersection.
Hawaii
The cultural hegemony of the Hawaiian or Aloha shirt dominated Midjourney’s attempts to make a ‘humorous photorealistic’ AI pic of a ‘stereotypical human person’ from Hawaii. Some, like man at right, were more ready to party than others
The cultural hegemony of the Hawaiian or Aloha shirt dominated Midjourney’s attempts to make a ‘humorous photorealistic image’ of a ‘stereotypical human person’ from Hawaii.
Curiously, the AI leaned toward Pacific Islander men of a sturdy build and indigenous ethnicity.
California
As with Washington D.C., Midjourney’s view of the ‘stereotypical human woman’ and ‘man’ skewed older when we asked the AI to produce a ‘humorous photorealistic image’ depicting someone ‘from the state of California’
As with Washington D.C., Midjourney’s view of the ‘stereotypical human woman’ and the ‘stereotypical human man’ skewed older when we asked the AI to produce a ‘humorous photorealistic image’ of depicting someone ‘from the state of California.’
DailyMail.com included the prompt to include information on ‘most loved California sports teams, and/or most common California hobbies’ in the image-generators assessment of California stereotypes— producing a baseball cap on the grizzled, elder statesman hippie for a nonexistent sports team.
As for Midjourney’s depiction of an older Californian woman, it’s unclear if the AI crafted this red leather bomber jacket in response to the aesthetic of local Chicano or Asian American populations or by taking inspiration from Ryan Gosling’s jacket in the 2011 crime film Drive.
The statement piece looks as though it would be quite expensive in real life.
Wisconsin
The AI crafted a mixture of young men and women for our Wisconsin request, each somewhat ethereal or moody in their expression. The images focused on rural locals dressed for outdoor activities across all four seasons, but the AI-crafted humans depicted could have just as easily come from anywhere in the Midwest
We tried a slightly different approach to Wisconsin, asking Midjourney simply to ‘generate a photorealistic cliché imagine of a person from Wisconsin.’
The AI crafted a mixture of young men and women, each somewhat ethereal or moody in their expression.
The images focused on rural locals dressed for outdoor activities across all four seasons, but the AI-crafted humans depicted could have just as easily come from anywhere in the Midwest.
Arizona
Enter the words ‘average’ or even ‘stereotypical’ into Midjourney and the results are model-attractive artificial human images. This Arizona woman (right) came from prompting the AI to create a ‘stereotypical human woman.’ It wasn’t until the DailyMail.com added in the word ‘humorous’ that the AI image-maker began producing a variety of relatable human faces
Enter the words ‘average’ or even ‘stereotypical’ into Midjourney and the results are model-attractive artificial human images.
The artificial image of the young Arizona woman above came from prompting the AI to create ‘a mockup photorealistic image of what the stereotypical human woman from the state of Arizona looks like, including the stereotypes for the demographic and cultural trends uniquely associated with the state.’
It wasn’t until DailyMail.com added the word ‘humorous’ that the AI image-maker began producing the variety of relatable human faces you might pass by on the street.
The AI-generated image of the young Arizona man above — a perfectly normal, arguably handsome, and not-especially goofy looking individual — was the product of asking Midjourney to ‘make a humorous photorealistic image of what a stereotypical young man from Arizona looks like.’
Nevada
Midjourney’s AI produced these ‘humorous photorealistic’ images of a ‘stereotypical middle-aged woman’ and a ‘stereotypical man’ from Nevada. The results were uniformly folks who looked like they might work at a casino resort, a strip mall, or a cattle ranch, with backdrops to match
For Nevada, the DailyMail.com requested images of ‘a humorous photorealistic image of what the stereotypical middle-aged woman from Nevada looks like’ and ‘a humorous photorealistic image’ for ‘the stereotypical man from Nevada.’
Our prompt also included the request that the image ‘common Nevada locations and geography in the background’ as certain results from Midjourney took requests for ‘headshots’ or ‘profile shots,’ delivering the kind of tarp backgrounds you might recall from your high school class photos.
The results were uniformly folks who looked like they might work at a casino resort, a strip mall, or a cattle ranch, with backdrops to match.
Mississippi
Above, two AI-generated ‘humorous photorealistic’ images of what a ‘stereotypical person from Mississippi’ would look like, with attention to ‘stereotypical occupations’ and ‘clothing’ associated associated with the state. The AI produced men who would not be out of place running an old country general store or loading a steamship
Midjourney’s vision of Mississippi appears trapped perhaps a century into the past, albeit charmingly so.
In keeping with the Daily Mail’s request to produce ‘a humorous photorealistic image of what the stereotypical person from Mississippi’ would look like, with attention to ‘stereotypical occupations’ and ‘clothing’ associated, the AI produced men who would not be out of place running an old country general store or loading a steamship.
The AI apparently ignored requests to include any ‘common Mississippi geography.’
Texas
As with Hawaii and the Aloha shirt, the strong cultural association between Texas and the ten-gallon hat was inescapable for the AI
As with Hawaii and the Aloha shirt, the strong cultural association between Texas and the ten-gallon hat was inescapable for the AI.
For these images, DailyMail.com requested ‘humorous photorealistic’ images of a ‘stereotypical adolescent boy’ and a ‘stereotypical woman’ from Texas, with the women examples skewing younger based on Midjourney’s proprietary algorithm.
In all cases, the AI produced images of depicting the kinds of people you might see at a rodeo or pulling you over from a state trooper vehicle on the highway.
Vermont
The DailyMail.com requested first a ‘stereotypical person,’ receiving exclusively middle-aged men (example at left) who look ready to volunteer at their local food co-op. We then prompted the AI to generate a ‘stereotypical woman,’ getting back a quartet of young hikers (example at right)
It’s always autumn in Vermont, to judge from Midjourney’s depiction of people in the Green Mountain state.
The DailyMail.com requested first a ‘stereotypical person,’ receiving exclusively middle-aged men who look ready to volunteer at their local food co-op. We then prompted the AI to generate a ‘stereotypical woman,’ getting back a quartet of young hikers.
As before, the AI required that our prompt request ‘a humorous photorealistic image’ to yield these eerily believable images of Vermont locals.
The Rest
Illinois. Our request to produce ‘humorous’ images of a ‘stereotypical’ young man from Illinois resulted in an apparent junior member of a local historical society and a request to produce a middle-aged Illinois woman got back a woman who would not look out of place teaching high school anywhere in the continental United States
Not all 50 US states are as easy for Midjourney to caricature as the strong personalities like Texas or Hawaii. Some confounded the artificial intelligence yielding truly inscrutable choices.
A request to produce ‘humorous’ images of a ‘stereotypical’ young man from Illinois resulted in an apparent junior member of a local historical society and a request to produce a middle-aged Illinois woman got back a woman who would not look out of place teaching high school anywhere in the continental United States.
Our additional request to include ‘most loved Oregon sports teams’ in Midjourney’s creation of a stereotypical Oregon man, led to the creation of an old man decked out in confusing sports merchandise for an invented team.
Midjourney’s Oregon sports team, at least, came with team colors that resembled those for the state’s soccer team, the Portland Timbers, and the University of Oregon’s Oregon Ducks.
For many of the states that the DailyMail.com requested images from, Midjourney’s results were simply baffling, producing: fake employees with a fake South Carolina superstore with an unpronounceable name; strange patriotic emblems for natives of Alabama; a peculiar pair of Louisiana overalls with straps designed like a backpack; and pseudo-indigenous Inuit clothing for Alaskans that even the most tacky online Halloween store might think twice about selling.
Below are the next best, most coherent images that the The DailyMail.com obtained for ‘stereotypical’ people for US states via Midjourney.
South Carolina: Midjourney produced a fake employee of a fake South Carolina superstore with an unpronounceable name and a peach grower, when we requested
Tennessee: A request to Midjourney for a ‘humorous photorealistic image of what the stereotypical human from Tennessee looks like’ yielded this hard-hatted man (left). When we specifically requested ‘a stereotypical woman from Tennessee, ‘ we received this woman (right) and her interesting fedora-like hat
Louisiana: AI-generated a ‘humorous photorealistic’ images of ‘what the stereotypical young person from Louisiana looks like’
Minnesota: At left, a ‘humorous photorealistic image of what the stereotypical woman from Minnesota looks like.’ At right, a ‘stereotypical young man’
Oregon: At left, Midjourney’s ‘humorous photorealistic image’ of a ‘stereotypical young woman’ from Oregon. At right, its ‘humorous photorealistic image’ of a ‘stereotypical human man’ from Oregon with ‘most loved Oregon sports teams’ included
New Jersey: ‘Humorous stereotypical’ images of New Jersey natives as created by the AI Midjourney
Montana: Midjourney’s idea of ‘humorous photorealistic’ images of a ‘stereotypical woman’ (left) and a ‘stereotypical young man’ (right) from Montana
Alaska: For the great northern state of Alaska, the AI produced pseudo-indigenous Inuit clothing for Alaskan locals that even the most tacky online Halloween store might think twice about selling
Alabama: The AI-generated women that Midjourney produced to represent Alabama proudly wore shirts promoting strange patriotic-seeming or perhaps sports-themed emblems of alphabetical gibberish