While everyone seems to be obsessed with making AI selfies and Studio Ghibli-style memes with the new ChatGPT 4o imaging engine (aka The Law of Average AI Posts), I’ve been experimenting with other things like infographics.
Step 1: I asked ChatGPT to list the 6 principles of sticky ideas from Heath and Heath. This is the framework I use in my NTU courses for evaluating the effectiveness and creativity of any content.
Step 2: I generated a 2D infographic with bright colours and icons. I also told ChatGPT how much text to include.
Step 3: Hmm, why not turn the 2D graphic into 3D? Bazinga!
However, as you can see, there is a minor error in the spelling of “Unexpected” in the 3D version. ChatGPT wouldn’t fix it, possibly due to the lack of space. The funny thing is, this is difficult to fix manually in Photoshop due to its 3D-ness.
More thoughts on the implications of multimodal imaging:
- PowerPoint slides just got supercharged. We’re no longer stuck to the incredibly boring SmartArt templates from the 1990s. Everyone can make fun infographics now, and not every infographic needs to be as precise as an Excel chart.
- AI-generated infographics will have an element of randomness and some minor errors, so every marketer needs to know how to fix these issues by prompting or through manual editing. You wouldn’t know it, but I extended the bottom of this image using Adobe Photoshop’s Generative Expand to give it more breathing space.
- Canva (PowerPoint for young people) must be sweating now. Its AI imaging features have always produced middling images and they acquired Leonardo.AI, which is another middling AI image engine.