I'm imagining the potential improvement since we've only just entered AI's infancy. With a "simpler" image prompt, such as a famous person eating shit, it could be more likely to produce a "perfect" image. Of course, it doesn't need to be a picture either. Deepfake video and voice generation can also apply.
The half joke is that there are always degenerate politicians, such that any gross claim has a chance to be true.