Intentional A.I. – Coaxing Images from Gemini and ChatGPT

Finding some Visuals

In a previous post, I mentioned my new Facebook page, “If you read nothing else today …” where I’ve been posting quotes and other information about some of the more interesting selections from my reading list. One of the challenges on Facebook is to catch people’s attention as they’re scrolling through the feed and text alone generally doesn’t do that so I’ve been trying to add some imagery to the quotes.

I decided to give A.I. a chance on this. As I’ve also mentioned before, I’m very selective in my use of A.I. – mostly I want to be sure I’m using it as an assistant and not handing off responsibility to it. Still, I thought this might be a good way to see what it could do. I’ve been getting into Jonathan Haidt’s “The Righteous Mind: Why Good People Are Divided by Politics and Religion” and some of the quotes are a little hard to put a single image to.

“Western philosophy has been worshipping reason and distrusting the passions for thousands of years. There’s a direct line running from Plato through Immanuel Kant to Lawrence Kohlberg. I’ll refer to this worshipful attitude throughout this book as the rationalist delusion. I call it a delusion because when a group of people make something sacred, the members of the cult lose the ability to think clearly about it. Morality binds and blinds. The true believers produce pious fantasies that don’t match reality, and at some point somebody comes along to knock the idol off its pedestal”

Jonathan Haidt, “The Righteous Mind”

I fed this into the free version of Google Gemini, which uses the Nano Banana engine to generate images, with the simple prompt “Can you create an image for the following quote …”. It seemed to prefer simply creating an image of a book with the quote on it although it wasn’t really able to manage the text that well as is often the case with A.I..

I asked it to try again and it gave me a couple of variations on this. Finally, I decided to be a little more specific.

Prompt: Can you create the image of reason as a crumbling idol thousands of years old being knocked off its pedestal?

Bingo! I like that it continued the context from the original quote by putting “The Rationalist Delusion” as an inscription on the statue, and got the spelling right. That’s the image I finally used on the FB page.

The next quote from the book reminded me of something George Carlin’s leeriness of the way people tend to organize.

“Many of us believe that we follow an inner moral compass, but the history of social psychology richly demonstrates that other people exert a powerful force, able to make cruelty seem acceptable and altruism seem embarrassing, without giving us any reasons or arguments.”

Jonathan Haidt

“The larger the group, the more toxic, the more of your beauty as an individual you have to surrender for the sake of group thought. And when you suspend your individual beauty you also give up a lot of your humanity. You will do things in the name of a group that you would never do on your own. Injuring, hurting, killing, drinking are all part of it, because you’ve lost your identity, because you now owe your allegiance to this thing that’s bigger than you are and that controls you.”

George Carlin

I plugged these two into Gemini and told it to leave the quotes off of the image because it had a serious problem getting them right. It came up with this:

Prompt: Try another image, without the quotes, and not quite as horror-inspired.

Not bad – it kept basically the same image but recognized that changing that one element would tone down the horror factor. Still, I decided to see what the free version of GPT could do and asked it to suggest an image first. You can see the full conversation at the following link:

https://chatgpt.com/share/699dd28e-3170-8000-bbc7-f233ffe95f65

Incidentally, I like that ChatGPT finally allows conversations with images to be shared via link. That wasn’t possible just a short time ago.

GPT actually suggested five different images, providing descriptions for each and how they related to the quotes:

1. The crowd casting a single shadow
2. A compass overridden by strings
3. Masks hanging on a wall, one missing
4. Birds flying in formation, one breaking away
5. A human figure dissolving into a crowd

Prompt: Can you generate #2 and #5?

I actually hoped for separate images but this was fine and not quite as dark as the Gemini offering.

One final exercise …

“The difference between can and must is the key to understanding the profound effects of self-interest on reasoning.”

“The social psychologist Tom Gilovich studies the cognitive mechanisms of strange beliefs. His simple formulation is that when we want to believe something, we ask ourselves, “Can I believe it?” Then (as Kuhn and Perkins found), we search for supporting evidence, and if we find even a single piece of pseudo-evidence, we can stop thinking. We now have permission to believe. We have a justification, in case anyone asks. In contrast, when we don’t want to believe something, we ask ourselves, “Must I believe it?” Then we search for contrary evidence, and if we find a single reason to doubt the claim, we can dismiss it. You only need one key to unlock the handcuffs of must.”

Jonathan Haidt

Again, I asked GPT to suggest images and specified that the colors shouldn’t be too grim like the previous images I’d used. GPT suggested 4 to 5 images for each quote. You can see the full conversation at the following link:

https://chatgpt.com/share/699dd4e4-1064-8000-b496-09f00faa1f53

GPT offered to supply the prompts before I asked for an actual image so I took it up on the offer to see what Gemini would do with it.

Original concept:

Visual Concept: “The Scale of Belief”

Main Image

A dramatic, minimal background (dark charcoal or muted neutral).

Centered in the frame:
A classic balance scale.

• Left side (slightly raised):
  A large stone engraved:
  “MUST I believe it?”
• Right side (heavily lowered):
  A small feather engraved:
  “Can I believe it?”

Even though the feather is lighter, the scale tilts toward it — symbolizing how little justification we require when we want to believe something.

Prompt:

A dramatic, high-contrast photograph of a classic brass balance scale centered in the frame. On the left plate sits a smooth pale stone engraved with the words “MUST I believe it?” On the right plate sits a delicate white feather engraved with “Can I believe it?” The scale is tilted clearly downward toward the feather, even though the stone appears heavier.

Background is a rich navy-to-indigo gradient with a soft violet glow, luminous but not dark or gloomy. A focused beam of warm light shines from above, illuminating the feather slightly more, with subtle floating dust particles in the light. Cinematic lighting, sharp detail, shallow depth of field, polished metal texture, emotional and symbolic composition, highly realistic, professional photography, 4:5 aspect ratio, optimized for social media.

Gemini did a pretty good job.

Ultimately, I used the GPT image because it seemed to have a warmer feeling to it.

Is this how I want to use A.I.?

I’ve noticed that I tend to allow myself to consider using A.I. in areas that my skills don’t cover. The idea of vibe coding or generating written materials is a hard pass but I’m willing to consider image generation because I’m not an artist. I use ElevenLabs to generate voice-overs from the scripts I personally write because my voice just doesn’t sound right on a video and it’s a lot faster.

On the other hand, there’s the question of usage and attribution. The images above are good and even remind me of the ones from the Bethel Bible Series course I took years ago. I’m not passing them off as artwork, though, and I label them as A.I.. They are solely for the purpose of lending visual flair and possible impact to words written by a human author. It’s similar to how I’ve also considered A.I.-generated instrumental music as a background for a video.

This use of A.I. doesn’t ask for anyone’s trust, unlike a vibe-coded app that could be riddled with security holes or an answer from a possibly hallucinating LLM.

It might sound like over-analysis to those who are more enthusiastic about adopting A.I. than I am but I think it’s important to keep asking these questions. New technologies don’t come with ethical guidance for how to use them. That will always be up to us and we need to base the analysis on more than productivity and profit.

For now, I’m good with this type of image generation although there’s a part of me that still feels it’s just too good to be true. This can’t stay free forever.