DALL-E 3 announced: https://openai.com/dall-e-3
The examples are for sure cherry-picked, so we will see, but they are WAY better than what I get from SD or DALL-E 2.
But most importantly, addressing all criticisms:
- Claims of no problem with fingers and spatial relationships.
- Claims of prompt engineering unnecessary.
- Integration with ChatGPT for "memory" and conversational generation.
- Safeguards against imitating the style of living artists and deep fakes.
- Training opt-out.
I think it was Emad from SD who said precisely that very soon, all those problems (legal, ethical, technical) will be solved and that the ugly reality of earlier versions was just a temporary intermediate step.
Again, I don't want to take what they say at face value; those are just claims. But I think it's all doable.
So, if the goal was not to have computer-assisted art generation happen, basing the criticism on those was pointless and futile. But if it was to correct those, they succeeded.