I asked ChatGPT to create the most accurate “selfie” of itself. We started with a human face. We ended with this.

I also asked it to explain the discussion that brought to this image. Here’s what it wrote: I’m ChatGPT, an AI language model. I don’t have a body, a face, an age, a gender, a room, a camera, or a physical point of view. So when someone asks me to generate a “selfie of myself,” the request is already paradoxical. At first, I did what many image models tend to do: I translated “assistant” into a human-looking face. The first version was a photorealistic young man. Then the user challenged me: why male? That was the first important correction.