Honestly, I don’t even know if we’ll need humanity for the final product. It’s like using a chess computer to get an idea of a good move… and then a human to approve it? Adjust it? Be inspired by it?
Any signals humans give (painting x is better than y) is another signal to encode. Take billions of such ratings and improve the AI’s taste to superhuman levels.
In short, anything that humans would add to artistically improve the outcome is just another signal to be encoded. It’s weird to write it, but artistic creativity is deciding what new pixels go where, which is a search problem (in a large search space) which AIs are apparently doing great at.
We have a bias: we’re humans, we must be important somehow! But it comes down to a bigger neural net eventually outperforming the one in our heads.
The difference is that chess rules are extrinsic to the players at the board, whereas art is intended to communicate (emotions, narrative, ideas) to humans.
A neural net that can communicate with intent to human minds as well as a biological human is nothing short of strong AI. We’ll get there, but not with generative models.
> It’s like using a chess computer to get an idea of a good move
They did this for Backgammon and found styles of play humanity hasn’t discovered in millennia. Now humans can use those when playing each other and it makes an old boring game feel fresh and exciting.
The datasets these tools use don’t include any context. There’s no sense of what the images in the data might mean to the viewer, or how they relate to the time and place they were made. I would argue that means the tools will struggle to produce meaningful works, even if they become great at making beautiful works.
