Learning To Work With AI

As I have developed EW Visual Arts, I have also been learning how to work with ChatGPT.

I use it to evaluate photographs, explore titles and captions, develop blog posts, prepare sharing layouts, and refine the website. Each part of the project has taught me something about what I need to bring to the conversation.

The most useful ingredient may be my willingness to ask why.

One day, a person I worked for as we were walking through the shop told me I asked the most annoying question: “Why?”

I replied, “How would you expect me to learn if I didn’t understand why?”

That memory came back to me as I considered how I work with ChatGPT. I want to understand the reasoning behind a suggestion. That helps me decide whether it belongs in my work—and learn something I can carry into the next decision.

An answer can sound convincing and still miss what I am trying to create. A caption can be beautifully written and still feel wrong beside the photograph. A website suggestion can make sense in general and serve my visitors poorly.

When that happens, I question it. Sometimes the answer needs changing. Sometimes I discover that I need to explain myself more clearly. Either way, the conversation becomes more useful.

With photographs, we have developed a staged process. ChatGPT first looks at the finished image without knowing my intention. It considers the visual hierarchy, balance, movement, relationships, framing, distractions, and what might sustain looking.

Only after that independent encounter do I explain what caught my attention, what I saw, and what I wanted to create.

That order matters. I want to learn what the photograph communicates before I explain it. If my intention and the first reading differ, we have something worth examining.

I also want recommendations that respect my Vision. My photographs may become painterly, simplified, abstract, or otherwise transformed through processing. Those choices help create the finished experience.

During the review of “This Way,” a painterly winter river scene, ChatGPT questioned whether the texture competed with the scene. I asked whether that response was influenced by the negativity surrounding photo editing.

That challenge exposed an assumption worth examining. The texture was visible, but its visibility alone did not establish a weakness. The evaluation needed to explain what it interfered with, or what it contributed.

I still wanted an honest judgment. Choosing an effect does not guarantee that it succeeds. I wanted the judgment to address the photograph I was making.

That conversation led to another adjustment. When a compositional dimension receives an assessment of three or below, I now want a practical recommendation alongside it. The suggestion should explain what a change might strengthen and what it could diminish. Retaining the photograph can also be a reasoned choice.

This gives me something to explore, compare, and decide.

The caption work has developed through similar questioning.

We settle a title, then explore Personal, Contemplative, and Open approaches. Comparing them helps me recognize where each directs attention. Sometimes a sentence sounds thoughtful but explains too much. Sometimes it repeats what the photograph and title already offer. Sometimes the simplest words feel closest to what I want to say.

I have learned to pay attention to those differences. I ask for another direction, shorten a sentence, reject an interpretation, or return to something I said casually in the conversation.

The words need to sound like me. They also need to leave room for you.

My hope is that a title and caption give you another reason to look at the photograph. Your response may be different from mine. That freedom is part of the work.

Blog writing gives these conversations more room. ChatGPT helps me develop a draft from what actually happened, and I review it for voice, emphasis, and accuracy. I want the history to distinguish what I intended, what I discovered later, and what was only suggested.

That has helped me separate stories about making a photograph from stories about the words beside it. Now, “Learning To See” offers a place for discoveries that emerge through evaluation and comparison.

There is a practical side to all of this, too.

For Pinterest, ChatGPT helped prepare layouts I could bring into Photoshop. That saved time and gave me a reusable starting point for presenting the work. I still choose the photograph and decide how it should be shared.

On the website, I have asked it to consider the experience of someone visiting for the first time. Can they find the work? Does the page give the photograph room? Is the print information clear?

When I wanted a search button, I described the experience I preferred. When testing showed that searches were finding keywords rather than photograph titles, I brought that result back into the conversation. We worked toward a change that I could test on my site.

That is another part of using AI: returning with what actually happened.