Perhaps some of you are wondering about my obsession with the pope; it’s simple practicality – I know there will be lots of pope images in the AI training models. Also, Marilyn Monroe.
The pope is also a good target to see if the AI can encompass odd requests. That was what got me started on my “… on a skateboard” kick. I was thinking that it’d be relatively easy for an AI to produce endless girly calendar pictures, but that some weird combinations of things the AI would not have seen before might be a good stress test.
Midjourney AI and mjr: “the pope drops fat tracks in st paul’s square and the crowd goes wild.”
Listen to me, “stress test” like I really care – I’m just having fun getting a computer to produce weird stuff. To me, this is all a part of gaming with computers, it’s a new game and it’s cool but perhaps I’ll get boring someday. I chose my words carefully, there, since with AI art, it will continue coming up with potentially interesting pictures as long as I can come up with creative and interesting prompts.
Cartier Bresson
A bit of art history: the great photographer Cartier Bresson used to speak of “the critical moment” – the moment at which something happened in a scene that made it interesting, and the photographer happened to be there and captured it (perfectly composed, exposed, and with the correct depth of field). I always thought that was bullshit, but it was useful bullshit because it encouraged artists to think in the moment and in the larger context simultaneously. Historically, it was bullshit, too, because nowadays we have cameras that are so high quality we can just shoot full motion, then pick the frame we like. I believe it was Ansel Adams who groused that the “decisive moment” in Cartier Bresson’s work was when he looked at a bunch of negatives on a light-table, and pointed and said, “that one.”
Similarly, when we are playing with AI art generators, the creative decisive moment is when we enter the prompt and hit return, followed by another decisive moment when we pick from the various options that the AI throws back at us. In other words, it’s all decisive moments. As you may recall, I have always been puzzled and fascinated by causality, which we perceive as a chain of events when it is actually a web – it’s all interconnected and each effect has an infinity of causes. It’s a purely human conceit (to be fair, we evolved this way) that we can pick apart the strands of the web and say, pace Aristotle, “this is the proximate cause.” The proximate cause was always “Cartier Bresson stood there with his camera and waited for one of trillions of decisive moments.”
Midjourney AI and mjr: “the decisive moment” AI primed with a screengrab from cyberpunk 2077 and the original Cartier Bresson photo
Aesthetically, by the way, I consider “cyberpunk” to be vapid. It appears that the entire aesthetic, rather than being anti-authoritarian and counter-culture, is in love with the mainstream of popular culture. Oh, you’ve got shit hanging out of your face and it glows in the dark? How … edgy. They just haven’t developed LEDs that small, yet, or yes, people would be getting glowing tattoos. The remaining question is whether they will be chromatophores derived from bacteria, or nano-machines, and whether there is any difference.
Back to the decisive moment: one of my favorite wildlife photographers, for a while, used to specialize in these gorgeous dynamic portraits of otters being all ottery. These were gorgeous, saturated, full-color, vibrantly detailed, photos. I was in awe until I learned that his method was to load a movie camera up with Fuji Velvia color film (notably punchy!) and shoot short clips of otters being ottery – then he’d get the film developed and pick the decisive frame and say “that one!” Nowadays you’d use a RED 1 on a drone and a digital editing station.
That’s how this shot was done: someone chose the exact right frame, taken just as his holiness lost his balance, panic-clutched the controls of his E-skateboard, and tore through the crowd with the swiss condottiere in hot pursuit. There were no injuries except to his surplice.
Midjourney AI and mjr: “the pope in st paul’s square showing off on his hoverboard.”
One final thought about “the decisive moment” – while I was writing this, I was casually punching stuff into Midjourney to see what came out the other side. Thinking about the cyberpunk aesthetic, I decided to ask midjourney to try a portrait of Molly from Neuromancer [No, Neal Stephenson did not invent cyberpunk in Snow Crash] and this was what hit my eye:
Thus, when we explore art, we explore ourselves. I had to look at that, and think, and finally admit to myself that I liked the lower right image more than the others because of the shape of the figure’s upper lip. Nothing else. I have written a bit about this, before, if I recall correctly, [stderr] Like other AIs, my preferences are strongly influenced by the weighting and sequence of my inputs, and if I am honest with myself, I am a much simpler machine than I thought I was. “Oh, boy, that gets me every time.”
Also, I think I will not do a whole posting of Marilyn tests, except that the V5 engine of Midjourney does a much better job with hands and, yes, skateboards.
Midjourney AI and mjr: “marilyn monroe in cream satin evening gown, riding on a skateboard in the style of hurrell”
If she were as square on to the camera as the other three, the one top right would be an acceptable passport photograph for a woman I was with in the mid nineties, to the point I had literal chills looking at it. Sounds like we both have a type and the AI’s idea of Molly is/can be close to it. Can’t say I’m surprised.
Gibson didn’t either. He famously didn’t want to watch Blade Runner while writing Neuromancer. And even Ridley Scott didn’t invent it – Philip K Dick and Mobius are closer to the source.
And THAT is the essence of AI art. I think that’s what (some of) the visual artists are bleating about – they’ve spent their lives getting good at rendering images, and now people who can’t do that, but have instead spent their lives getting good with the words, are now able to translate THAT skill into visual images directly. It’s still definitely art, but then I’m a word guy, so I would say that.
@sonofrojblake: testing, testing… I bet you like the art of Milo Manara, particularly how he draws faces.
(Edit: you may notice I actually was careful to dodge the question of who created cyberpunk, in my phrasing. It ain’t Neal, is all.)
I always thought Molly was at least part Asian. And none of those renderings show the mirror-eye enhancements.
Youtube is currently filling up with AI renders, which is going to be … interesting. Will the algorithm down-play them, or will they crowd out everything else?
I am not a non-fan; I simply do not know. I will say that traditional artists felt similar conflict when photography studios began to pop up all over the world, and what was formerly a difficult and time-consuming creative process became a difficult and less time-consuming creative/mechanical process.
A have few vices (ha! actually, I have at least six bolted to various posts, benches, etc) but one recent addiction is the ballroom-style “balenciaga” meme-videos. They are not quite good enough yet but by next year they will be overplayed, and amazing. The idea is simple: take characters from star wars or whatever and have the AI dress them in the style of Balenciaga. If I were Balenciaga, I’d be worried because the AI do Balenciaga better than Balenciaga. (e.g.: the papal reactive armor in my earlier post)
Closing in on Molly:
Here’s the prompt. I may have to feed some sunglass images in to trigger mirrorshades
Edit: oops lost a finger! Sometimes must add “no missing fingers, no extra fingers” to the prompt.
Lots of echoes of Andrew Eldrich in his prime… Or Pat Benatar. Nothing wrong with Pat.
By the way, you can do the equivalent of “image2image” by sending midjourney its own output and specifying changes to it. i.e.: earlier images from itself + “mirror eyes::5”
Raging Bee:
I always thought Molly was at least part Asian. And none of those renderings show the mirror-eye enhancements.
One of the amazing stylisms Gibson pulls off so well is his deliberate vagueness about practically everything. I don’t think we are ever explicitly told she is part asian. She has high cheekbones, shaggy black hair, and, um… The mirrorshades and burgundy nails. Otherwise she dresses “like a street samurai” uh, what? More vagueness. Gibson’s so evocative that our mind takes over and provides its own imagery. You know, kind of like an AI does.
Compare to quote from ‘Repo Man’
MILLER: A lot of people don’t realize what’s going on. They view life as a bunch of unconnected incidents and things. They don’t realize that there’s this like, lattice of coincidence that lays on top of everything. I’ll give you an example. Show you what I mean. Suppose you’re thinking about a plate of shrimp. Suddenly, somebody’ll say like, plate, or shrimp, or plate of shrimp. Out of the blue. No explanation. No point looking for one either. It’s all part of the cosmic unconsciousness.
OTTO: Did you do a lot of acid, Miller? Back in the hippie days?
Images 2 and 3: I would judge that Midjourney has failed to convey a ‘decisive moment’ in the manner of Bresson. In image 2, the moment captured will be followed by the foot hitting the water, resulting in ripples and eventually a soaking. In image 3, the step will be followed by another step. Capturing a reflection off a puddle just wasn’t the active element to focus on.
@6: We need Midjourney responses to ‘street samurai.’
In the book The Godless Constitution by R. Laurence Moore and Isaac Kramnick, if you pay attention you may notice the phrase “dime-store atheist.” What does it even mean?
In other words you like a pronounced philtrum, shared just because I love the word.
I would agree that the lower right Molly is the most attractive version, but it’s more than just her upper lip. Adding the widows peak to her hairline makes her face heart shaped, and her eyes are almost freakishly large and widely spaced. (Like an infant, though I can’t remember the technical term for this at the moment). The slightly open mouth is suggestive, and markedly different from the rather grim, grimy facial expressions on the other three.
Those tiny changes (for women) are highly attractive to a large proportion of humans.
High cheekbones and well defined lips are what is generally termed ‘good bone structure’.
I see Jean Harlow on a skateboard, possibly because it’s Harrell style, but the AI has definitely gotten much better at hands.
I’m not sure about decisive moments. I think dynamism might be what is actually captured in that image, and is exactly what makes the image compelling. There is a story in what is about to happen, which creates an automatic internal dialog between the image and the imagination of the viewer.
Is he about to be knee deep in a pothole, or will he just ruin his shoes?
Jazzlet@#10:
In other words you like a pronounced philtrum, shared just because I love the word.
I do, but it’s more than that. I like sharply defined lips, with the upper lip a cupid’s bow. And philtrum, natch.
That was a word I collected in high school, playing Boggle, along with another favorite, the aglet.
When I was a kid my parents had a copy of Willard Espy’s Words at Play which was kind of a literary almanac of sorts. Lots of poetry, weird words, etc. I learned animadvert and usufruct from that book, and managed to use animadvert, once, when I was flirting in college. Other than that, my flirting was a failure.
@Tethys,11: “neoteny”. Last saw the term applied to Jenna Coleman.
“my flirting was a failure”
I disagree. You successfully weeded out the ones not impressed by what you offer. It’s a common misconception that “successful” pickup game means being able to pick up anyone, when in fact it’s important to recognise asap when you’re wasting your time and theirs and withdraw politely.
In addition to neoteny (or neotenous), there’s also pedomorphic (or paedomorphic)
Maybe you could get mirroreyes (or something close to that) by using “mirror sclera lens” (maybe add “contact” before lens) as an input prompt?
You know, it looks like no-one has a mirrorshades or reflective sclera lens available for sale, as best I can tell with an image search. The closest to something like it is “blind white” (which looks translucent in the center)
It may be that making a mirror scleral lens is incompatible with making a lens safe to wear.
The Holy Rollers!
He should wear a helmet. And if those robes catch in the drive, he’ll have a bad day.