In this week’s newsletter on how to spot AI-generated text, I wanted to use a four panel comic to illustrate a point from this paragraph:
“You’ll also notice AI’s obsession with paired adjectives (“unique and intense,” “highly original and impressive”) and its preference for connecting simple statements rather than crafting complex thoughts. The sentences are technically correct but mechanically rhythmic, like reading a well-programmed assembly line. Which…
…well yeah, which it is.”
Here’s the prompt that I initially used with Sora:
A four panel comic of a robot and a human. Both are looking off-panel at something in the non-visible distance. The robot is describing it:
Panel 1: Robot: “It’s unique and intense..”
Panel 2: Robot: “Furthermore it’s highly original and impressive.”
Panel 3: Both are silent.
Panel 4: Human: “Yeah, it feels good to me.”
I set it to run two generations, and you can see the end result here:
In this first generation, the last panel in the left comic changes the robot to a human. Otherwise, spot on. In the second generation on the right, the robot does all the talking.
So I remixed it, this time by highlighting the extra human and telling Sora to “Add the robot here, looking at the human.” And here are those results…
It’s like Sora has poor impulse control or something. It can’t not add something in these images!
In the end I Photoshopped the last panel in the first generation just to make it easier.
