You can use Stable Diffusion to make more specific compositions, but you usually need multiple steps and a little bit of manual editing.

It's definitely possible to get concrete representation out of DALL•E, although it takes some experimentation to achieve realism within the constraints of the tech today. Still, we're not far off at all from indistinguishable synth imagery.

Some generated work in a photographic style: — https://www.hearthe.art/ — https://objkt.com/profile/heartheart/created

i think people underestimate just how many commercial images effectively fall within the 'ambience' category. i'm really convinced that in the mid to longterm, the only viable future for human artists will be a matter of GANs being subjected to the same sort of social stigma that is attached to other forms of highly economically lucrative, but morally dubious practices like child labour. ultimately sustaining a living in art has always been a matter of convincing the layman that what we produce has value and this will be the same at scale. if society does not value human-made art, there wont be any.

Self-driving cars have already been not ready for prime time for 15 years and they have at least 5 more years in this state.

But self-driving cars have a serious good enough threshold that isn't remotely as life-or-death for illustration