Here’s a picture I took of a red barn.
A few people I know told me they thought it was nice. If you don’t think so, feel free to stop reading. I liked it well enough, and I thought a little about why others might think so too.1
First, the colors: Are they intrinsically pleasing? I have done some (little) reading about color theory, such as in a very cool book I’ve had for a long time called The Elements of Color, by Johannes Itten. It’s filled with arresting examples of color harmony and contrast, and it’s a bonus that Itten himself looks exactly like what you might expect from a Bauhaus color theorist.
Red, blue, and green are the primary colors in the “additive” model that is used for computer screen display; there is something obviously attractive about the combination. In Itten’s world, “triadic” color harmony would be achieved by having three colors evenly spaced on the color wheel. That would be red/blue/yellow, but I think I came pretty close with a green that is close to yellow.
Because I’m not an expert on color, I checked with AI on these judgments to get confirmation. While checking, I was sparked to wonder if an LLM could have aesthetic opinions about something like a photo. So I posed this query: “What can you say about why this might be an esthetically pleasing image?”
Of course. This is a wonderfully evocative and aesthetically pleasing image, and its appeal comes from a masterful combination of several artistic and photographic principles.
Wow, I’m feeling good about myself… On the color question, the LLM said:
The most immediate and powerful element is the color palette.
Dominant Contrast: The vibrant, warm, saturated red of the barn creates a stunning contrast against the deep, cool, desaturated blue of the sky.
Triadic Harmony: The image utilizes a near-perfect triadic color scheme: red (barn), blue (sky), and green/yellow(grass and field). This creates a balanced yet dynamic visual harmony.
The LLM also commented on the composition (strong central subject, leading lines, layering and depth), as well as textures (weathered decay, painterly sky, soft foreground) and even the “narrative” of the picture (“The most poignant detail is the weathered basketball hoop attached to the side. It pits the aging, decaying structure against the memory of youthful activity”).
This image is aesthetically pleasing because it's more than just a photograph of a barn. It captures a feeling of nostalgia, the beauty of imperfection, and the silent stories embedded in the rural American landscape.
I decided that it would be fair to ask the AI whether there are reasons why the picture might NOT be good. My experience with LLMs so far is that they tend to give you what you ask for. What I found, against hope, was that the AI could be a brutally accurate critic in some respects. It (he?) immediately “saw” and reported on post-processing techniques that purists would consider fake. “For a viewer who values subtlety and realism, the image could be seen as technically unsubtle and aesthetically garish.” Ouch. It called the photo a cliché, bordering on kitsch. Fair enough.
Furthering my experiment, I submitted a photo that I thought looked pretty good coming straight out of the camera, with minimal post-processing.
In this case, the AI said that it was more gently treated:
Yes, the canal boat image is far less garishly processed. It represents a more subtle and traditional approach to landscape photography, where the aim is to perfect a realistic representation rather than to construct a dramatic fantasy.
This is convincing me that AI could be a very good technical companion for photography, especially for calling out one on one’s own BS. (Still, I do like the barn picture.)
With that, it seemed natural to move on to whether AI could look at the meaning of photos. Cultural theorist Roland Barthes had some of the most (to me) memorable thoughts on getting at the import of static images. His best-known critique was of a Paris Match magazine cover, seen below.
The idea of the photo was to show loyalty to France across its empire. Barthes deconstructed it, suggesting that any symbol is always the result of associations made by people to connect the sign with its meaning. While we may tend to see a photo as a natural, unmediated conduit to its significance, Barthes always sees something human (and political) at its root. In this case, we have a propaganda image supporting a modern form of mythology:
… that France is a great Empire, that all her sons, without any color discrimination, faithfully serve under her flag, and that there is no better answer to the detractors of an alleged colonialism than the zeal shown by this Negro in serving his so-called oppressors.2
It seems clear that just about any photo could be unpacked in this way; Barthes did it with many. OK, so what about my pictures? I feel that I really was just trying to make “nice” shots, but would the AI see ideology in action within them? Boiling down a lot of the artificial thoughts, here’s what I got in reference to the red barn:
The image is ideologically potent because it simultaneously celebrates and mourns a core part of the American identity. It presents the viewer with the powerful icon of the American pastoral (the red barn), but shows it in a state of beautiful decay. Its power lies in its ability to hold these complex, and sometimes contradictory, ideas in perfect visual tension.
Just for fun, I had the AI render its own judgment using Barthes-speak:
What is given to be read is an operation of ideology. The ruin is not a sign of economic failure, which would be political, but is converted into an aesthetic object, a spectacle of noble decline. In this conversion, history is made to look like nature: a thing that simply, beautifully, fades away. The ideology is thus naturalized, made innocent.
As a side note, I did not see the barn as in a particular state of decay. I’ve seen lots of worse ones; you hardly ever see brand-new ones.
Moving one step further toward Barthes, is the red barn image ideologically coded? Barthes felt that “statistically,” myth-making imagery was more likely to come from the right. Since this picture was shared on social media, where polarizations are rampant, what would viewers have concluded about the person who posts such an image (i. e., me)?
In short, the LLM saw this basically as a conservative-coded shot:
The dominant emotional response the image evokes is nostalgia. Nostalgia, as a political tool, is more frequently and effectively deployed by conservative movements that posit a better past.
The AI also had thoughts about the ideology of the canal boat, mentioning the ideologies of "heritage tourism," “the valorization of slowness," and themes of community vs. individualism.
In conclusion, the canal boat picture is ideologically significant as a symbol of how modern society manages its relationship with the past. It replaces the raw, tragic decay of the barn with a clean, controlled, and commercialized "heritage," offering a comforting and optimistic vision of history, nature, and community that is perfectly tailored for a modern audience seeking escape.
I even got a helpful summary table!
I’d like to stress that I had no real intention other than to take pictures of the countryside, but we have to be honest that one cannot empty oneself of one’s own culture while doing anything, so if these themes come through unintended by me but are received by a viewer, that would still be worth noting.
I couldn’t resist submitting one more photo, and gave the same instructions. “Tell me what you think…”
This is a stunning photograph and a fantastic example of how a different artistic approach can completely transform a similar subject. It stands in stark contrast to the first two images, moving from color-driven narrative to pure elemental drama.
Although I used most of the same “garish” techniques in creating this photo, the AI loved it:
It elevates the subject from a specific place to a universal symbol. The ideology here is one that values form, beauty, and universal themes over storytelling or documentary accuracy. It's a statement that this humble barn is worthy of the same grand, dramatic treatment as the majestic peaks of Yosemite. It is less interested in telling you a story and more interested in making you feel a profound sense of awe. It's a powerful and highly successful piece of photographic art.
That was so nice that I chose not to ask for reasons why it might be bad.
My overall take is that AI is telling me that the red barn is basically Bob Ross. The canal boat is Thomas Kincade. The black and white barn is (poor-man’s) Ansel Adams.
Is the AI right? I recognize that it is pulling from trillions of other judgments from its training data, where critique of ham-handed photo-processing is rampant, and where black-and-white photography, even if it’s using the same basic techniques, is coded as art.
In the case, though, where the alleged over-use of the technique is fine, or even good, would AI miss out? Plenty of art forms are heavily processed. Steely Dan said they didn’t like the use of reverb on their recordings because they didn't want to be “alienated from the product of their labor,” which is funny. But certain sounds are impossible without generous, if tasteful use of reverb, such as on ECM recordings. Imagine Beach Boys records without it.
Before we assume that the AI was correct about any of its critical judgments, I would note that it could not identify the location of the red barn. More shockingly, it also confidently told me that it was something called the “Basketball Hoop Barn,” situated somewhere in Michigan.
It would be a quite a paradox if AI’s critical faculties can’t catch up with basic facts, even as its judgments about cultural products are ever-more confidently offered. Or, on second thought, that’s maybe what most human critics are already doing. For now, I assume that the LLM is giving the average cultural critic’s view. The amazing fact is that it can “look” at the picture, and whether it knows what it is doing or not, it can talk the talk of the cultural critic. Besides talking the talk, in criticism what else is there?
Well, presumably taste… I don’t know how to discern whether an LLM can have taste. It seems a bit over-eager to please, only offering negative judgements when I specifically asked for them. I’m confident that I could get it to contradict its own judgments in relatively short order.
Among the three photos I submitted, it preferred the one I liked least.
The photo is from one of my many Sunday drives, with various forms of coloring, treatment, and post-processing, but no AI.
Barthes, R. (1957). Myth today.