I am very confident the image at the top of this article is generated by any of these currently popular text-to-image models. This feels like the beginning of ignoring artworks I see online, foreshadowed by header images at the start of tech blogs.
But at the current state of things, the details are usually wrong in every subject, not only people.