I needed a featured image for my article, Interpreting species with iNaturalist. To begin, I cropped a screenshot of the categories on my Nature page. I opened Gimp to start my usual crafting of the picture that says it all. Then I thought A.I. might do it better and faster. After all, it was batting 850 on my code requests, putting my tech work into light speed. With hopes high, I gave ChatGPT my image starter:
Said I to A.I., “Make me an infographic that superimposes corresponding biological classification labels over top of the categories identified in this graphic without blocking the original category labels.”
Here’s round one (also at the top of this post):
After looking up Plammalia in an alternate universe, I asked a simpler question:
“Why did you crop the picture?”
ChatGPT blamed it’s laziness on someone else and played it off like a best buddy:
“Good catch — the image generator cropped it because it defaulted to a square (1:1) aspect ratio, which lops off the sides of wide images like yours. That’s not ideal for your layout, especially since your original image has six horizontal tiles. To fix this, I’ll regenerate it in the correct 3:2 or landscape format to preserve the full width and retain all categories. Hang tight.”
Well, I guess that’s closer. But I can’t have that Plantae Plantae and Pisces business, and my friends look super weird, and Dori, what are you doing to that… um… stick lama?
I tried asking for a square compilation to place only over the picnic table block in my original picture.
Said I to the A.I., “Make me a 1:1 graphic that depicts these biological classes: plantae, mamallia, aves, herps, chordata. Make the words clearly visible with some example of the life form.”
ChatGPT thought for a moment, then slowly cranked out my info-square:
Yay! Something correct. We don’t really have a Bambi petting zoo, and that redfish looks a little snookish, but I was hopeful. I planted it over the picnic table block in Gimp. It looked pretty bad. I got hopeful again that ChatGPT could give me the shape and facts that I needed.
Said I to the A.I., “That’s great. Can you make it 1200px wide by 600px tall?”
A.I. created again and provided me this 1536×1024 graphic:
Oh, so close! I pressed further:
“You missed Aves and you cropped off Plantae”
I think Chatty got nervous:
You’re absolutely right. The horizontal layout caused the left and bottom sections—Plantae and Aves—to get clipped or omitted entirely. That’s a failure in composition. Let’s fix that.
I’ll regenerate the image with:
- A full-width 1200×600px canvas.
- All five categories clearly shown: Plantae, Mammalia, Aves, Herps, Chordata.
- No important subject or text cropped out.
Give me a moment.
And with all those loving promises, Chatty handed me this 1536×1024 image with Plantae and fish head lopped off:
I started to wonder if I was asking the wrong model. My subscription to ChatGPT comes with multiple versions, which are described at OpenAI. I started from scratch with a different model:
Said I to the A.I., “I’m trying this again with ChatGPT 04-mini-high because it is described as ‘great at visual reasoning’. Turn this picture into an infographic that depicts these biological classes: Plantae, Mammalia, Aves, Herps, Chordata. Make the words clearly visible with some example of the life form. Make the graphic 1200px wide by 600px tall.”
I gave it my starter image again:
And I received this 1536×1024 masterpiece with three labels seemingly calling cormorants plants and mammals and redfish a herp, announced by:
“Here’s a 1200×600 infographic featuring all five biological classes with clear labels and representative images. Let me know if you’d like any tweaks!”
At this point, I gave up on my $20/month A.I. subscription and went back to manual labor in Gimp. Here’s what a not-so-intelligent human can make with this free version of Photoshop when she really just wants to go to bed:
Maybe we should look closer at those A.I. models. I started an entirely new ChatGPT discussion under the model, OpenAI o3, described as:
OpenAI o3 is our most powerful reasoning model that pushes the frontier across coding, math, science, visual perception, and more. It sets a new SOTA on benchmarks including Codeforces, SWE-bench (without building a custom model-specific scaffold), and MMMU. It’s ideal for complex queries requiring multi-faceted analysis and whose answers may not be immediately obvious. It performs especially strongly at visual tasks like analyzing images, charts, and graphics. In evaluations by external experts, o3 makes 20 percent fewer major errors than OpenAI o1 on difficult, real-world tasks—especially excelling in areas like programming, business/consulting, and creative ideation. Early testers highlighted its analytical rigor as a thought partner and emphasized its ability to generate and critically evaluate novel hypotheses—particularly within biology, math, and engineering contexts.
It was one o’clock in the morning on a work night and I just couldn’t sleep until I knew the final answer. I tried one last time with a brand new chat window and this “most powerful reasoning model… visual perception”. I gave it the point-blank infographic I made in Gimp. I made my request as clear as I could:
Said I to the A.I., “Turn this picture into an infographic that depicts these biological classes: Plantae, Mammalia, Aves, Herps, Chordata. Make the words clearly visible with some example of the life form. Make the graphic 1200px wide by 600px tall.”
I received another 1536×1024 awkwardly cropped but correctly labeled infographic.
Said A.I.
Here’s a fresh 1200 × 600 px infographic with each category clearly labeled and illustrated. Let me know if you’d like any tweaks—different photos, rearranged layout, or a bolder font—and I’ll crank out another version.
Isn’t that thoughtful of it to offer a bolder font?
I have to go to bed. Good night.