"As AI progresses, it’s less and less “AI.” (Lev Manovich)

Just posting a very interesting comment by Lev Manovich (Professor at City University of New York - http://manovich.net/) re AI. He has been experimenting with Midjourney for a while to generate quite beautiful artworks. He made an interesting post on LinkedIn today (quote below):

https://www.linkedin.com/feed/update/urn:li:activity:7151453175592505344/

"Paradoxically, as image AI generators advance, they become less (as opposed to more) “authors” of the works. They become “less AI” and more like regular tools that obey our detailed instructions. With early versions of Midjourney (3 and 4) it was inventing all kinds of strange things, misinterpreting your intention and generating strange, alien and often very poetic imagery. V 5 imposed more realism but I was still able to confuse it by using multiple images as part of a prompt. And now V 6 really wants to obey you. You have to tell it everything.

So now it’s is less about “AI authorship” and it’s charming alien ideas - and more about banality of human imagination and very predictable and boring human “creativity” and suffocating “beauty” masses love.

In summary: As AI progresses, it’s less and less “AI.”

And here is the answer about the usual question about “authorship.” AI was more of an author in early Midjourney versions and less so in new versions. It visualizes what you want, yes, but does not add unusual twists and strange ideas."

I find this quite a compelling idea...

Lev is also writing and publishing a book (with Emanuele Arielli) on "Artificial Aesthetics: Generative AI, art and visual media" - one chapter at a time. You can download all the completed chapters here:

http://manovich.net/index.php/projects/artificial-aesthetics

© Body text Adam Hyde, 2023, CC-BY-SA, excerpts ©Lev Manovich
Image public domain, created by MidJourney from prompts by Adam.