As I noted there, I am not in any way opposed to the technology--it's amazing. But the way it's currently being employed is basically IP theft.  All of these AI generators required training data in order to work at all. Every single online platform (Stable Diffusion, Dall-E, Midjourney, etc.) uses "scraped" data obtained from the internet. This becomes evident when mutated versions of the Getty Images or Shutterstock watermark sometimes appear in the generated images.

You can't copyright a style (why you can actually see McQuarrie and Ryan Church and Doug Chiang in those Star Wars if it starred Mick Jagger and Brigitte Bardot generated images) but that doesn't mean the exact copyrighted data wasn't used without permission. 

Things get worse with voice-over artists.  There was a recent mod for Skyrim that took spoken input from the user (via microphone), ran it through a voice to text generator, ran the text through ChatGPT which generated a lore-friendly response.  So far so good, but they then took that text and ran in through a text to voice generator using the character's voice actor's voice!  Having full actual in-character conversations with even minor characters is now a thing!  Mind boggling.  Unfortunately, the voice actor doesn't get paid for this.

I also noted this can all be done ethically where everyone wins.  For instance, you could pay the VO actor twice as much for the scripted material and a training block where they can say things in a nuanced fashion to improve the training (right now, it's pretty deadpan). Likewise, you could pay Shutterstock for training block data, or if you have deep pockets like Disney, compile their own, where volunteers get paid $20 or whatever to have their picture added to the training block under ideal lighting conditions.

Considering how eager some corporate types are to not pay artists (see the current writer's strike), this all has to be addressed before it's out of control.  For instance, AI Comics may be laughably bad now, but in 10 years it could put every artist and writer of every stripe out of work, because (button click) and voila! Perfect stylistic facsimile brand new Claremont/Byrne/Austin comics that they never made.  Took a couple minutes.  They could generate them on demand in seconds.  Or CG/AI movies with a perfect facsimile of Bogart, etc. All of it based on stolen art from the past.

The most depressing evolution I've seen of this recently is the idea that the composition of a song shouldn't be protected under current copyright since all music is derivative (whereas somehow the recorded performance of the song does earn money for the recording artists, though not the writer).  What would replace it would be some kind of creative commons fund that people pay into.  It'd be one thing if we lived in a moneyless utopia where artists didn't need to eat, but it all seems pretty fucked in the context of a capitalist society.  Check a highly biased* video essay about it HERE.

It's a pretty stark reminder of the generational divide driving this conversation.

*I suspect the vlogger favors this approach because he's an improvisatory performer.  In my experience, jazz players don't place a huge premium on the importance of composition. An understandable perspective but also an unfortunate one.

It's the first AI art that I find significant, because it's conceptually right. It's not using it as a shortcut or to not pay actual illustrators/designers, obviously Marvel has the money.

The award winning animation studio actually pitched the idea to Marvel, not the other way around. And their artists are credited and worked for months on this, and it shows.

It looks bad, like a couple of generations apart from actual versions of Midjourney et all (say, what Midjourney delivered 6 months ago?), and it's not trying to hide it.

And that's the whole point. 

For it to look unsettling, blurred faces, generic "AI Look" so to speak. Perfect for a show about aliens trying to imitate humans, and seizing the whole AI debate in the mainstream and how weird we all feel about it.

AI *has* a style. And in this intro they distilled it perfectly IMHO, and got the desired effect.

The episode itself, I found generic. 

The intro, that was something.

(And it would make no sense to have "actual artists" try and imitate the AI look as I've seen a lot of people argue, then the whole point would be lost).

I actually think it's a critique of AI really, not an endorsement. 

Remember, the Skulls are the bad guys.

As I said, this can be done ethically, they just didn't.  At the moment, you really can't.  Stable Diffusion having a 3 month opt-out program for artists is hilariously inadequate. I for one didn't even hear about it until long after the opt-out program was over.

This isn't like Alpha Zero inventing it's own chess theory by playing against only itself millions of times and generating it's own neural network by the experience of playing those games.  AI Art generators require the internet and billions of human-made images on it (much copyrighted) to train with in order to work at all.