It just so happens that the image (Global Migration to cities) touched on a different topic that I have some interest, namely the limits of AI. ChatGPT and DALL-E have been trained on textual and 2D images to such a high level that they can reason within those frameworks for example this by Lessig: ChatGPT, or: How I Learned to Stop Worrying and Love AI. I left response to that effect: https://medium.com/p/b84af455022f

But apparently artificial neural networks (ANNs) are still lacking because they are not fully trained in the function, physical qualities and physical inter-relationships of the depicted items. Basically DALL-E constructed a simple 2-D image from specified items, but did not “take a photo” from a simulated reality. A human artist would imagine a reality and construct the image. Even then, of course, the image can have shading/shadows which are inconsistent with physics etc. Video games do a better job with this.

For now, you are using ChatGPT and DALL-E as tools just as other human artists have used paint and canvas. At what level of training/artificial experience, will the ANNs then be able to imagine a reality such that it can then create a “camera image”? I think it will still be short of becoming an intelligent organism, but it will be very close.