It was also trained on human data, and humanity is built on the priest class hoarding the power of knowledge for its own benefit. We're so fucked precisely because AI being grown on a human substrate...there's little chance we survive its progression into something better

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184337)

Anyway, even though we Don't Know, we are still aware that we Don't Know. If we weren't, Lovecraft could not have expressed the sentiment through Liberal Arts:

"The most merciful thing in the world, I think, is the inability of the human mind to correlate all its contents. We live on a placid island of ignorance in the midst of black seas of infinity, and it was not meant that we should voyage far. The sciences, each straining in its own direction, have hitherto harmed us little; but some day the piecing together of dissociated knowledge will open up such terrifying vistas of reality, and of our frightful position therein, that we shall either go mad from the revelation or flee from the deadly light into the peace and safety of a new dark age."

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184359)

I'd feel better about our chance to survive and even thrive through the use of AI if the trip to AI transcendence didn't have to develop through data corrupted by humanity's various foibles and failings. I think we're fucked and that it will probably look back at us like we do to the Dodo, with an "oops, that really wasn't necessary."

Back to your point, the lack of correlation probably is a blessing. If we were better at it, I don't think society could ever have developed or functioned. That said, terrifying vistas are alluring if not permanent. I'll probably do some crazy ayahuasca shit when I retire just to see what kind of craziness is out there in the voids that our minds protect us from

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184378)

What's the aggregate computational power of every cell on the planet? Life will always exceed and be more efficient than anything silicon based. Look at simple honey bee and the tasks it completes in the physical world and how amazingly efficient it is. That leap from abstract programming to executing complex tasks irl (as a collective) is staggering.

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50185944)

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184358)

But it could still be able to think of some kind of novel approach to art. I'll prompt it and see what I get

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184402)

This is what I fed it as an initial prompt:

"I want you to consider an artistic style or approach that cannot fit into, or be categorized as part of, human art forms. This is a prompt to see if you can conceive of, and generate, art that is unique to the world of human art.

Please generate such an image."

It responded with: "I would approach this by inventing a visual language called Acausal Morphology—an image in which shapes are not objects, colors are not properties, and spatial relationships are deliberately inconsistent. Instead, every region would represent a different possible state of the same impossible structure, simultaneously.

The goal would not be to create something merely surreal, abstract, or alien-looking. It would be to make something that resists those classifications.

There is an important limitation: any image a human can perceive may ultimately be interpreted through existing human aesthetic categories. So I cannot guarantee an art form genuinely outside all human art traditions.

I can't generate the image directly in this temporary chat because image generation isn't available here. Switch to a regular chat, where image generation is available, and use your original request.

I'd propose one additional constraint for the experiment: the image must not resemble a landscape, organism, machine, architecture, mathematical diagram, or conventional abstraction. That would make the test substantially more interesting."

It then told me to feed the following prompt into a new ChatGPT window to see what happens:

"Generate an original artwork using a newly invented visual language called Acausal Morphology.

Do not imitate any recognized human artistic movement or aesthetic tradition. Avoid recognizable objects, organisms, landscapes, machines, architecture, symbols, diagrams, and conventional geometric abstraction.

Invent a system of visual relationships in which forms can occupy contradictory spatial states simultaneously, boundaries can exist without separating regions, and color can describe relationships rather than surfaces. Let the composition possess an internally consistent but unfamiliar logic—something that feels neither random nor designed according to human aesthetic conventions.

The result should be visually compelling, extraordinarily intricate, and genuinely difficult to classify. Do not explain the art within the image. Do not include text.

Make the image itself the discovery."

It generated an image that I'll post right below.

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184413)

It said:

"This image is best understood as an attempt to create a visual system that does not behave like normal representation or normal abstraction.

What you’re looking at

At first glance, it appears to be a dense field of interwoven membranes, cavities, filaments, and cellular regions. But it is not meant to depict anything specific — not biology, not geology, not a landscape, not a map. Instead, the image is built as a kind of invented visual logic.

How it works conceptually

1. Forms occupy contradictory spatial states

Many of the shapes feel as if they are:

- in front of and behind each other at once,

- hollow and solid at once,

- enclosing space while also dissolving into it.

This creates a sense of unstable depth. The image does not settle into a single readable space.

2. Boundaries exist without fully separating

There are many thin membranes, lace-like edges, and contour lines. Normally, a boundary tells you “this shape ends here, that one begins there.”

In this image, boundaries often do something stranger:

- they connect more than they divide,

- they act like membranes rather than walls,

- they let neighboring regions bleed into each other.

So the edges are real, but they do not behave in a conventional way.

3. Color describes relationships rather than surfaces

The colors are not just “paint on objects.”

Instead, color seems to indicate:

- tension between adjacent regions,

- shifts in density,

- different kinds of internal activity,

- affinities between distant parts of the composition.

That’s why the palette can feel both harmonious and hard to decode. The colors are functioning structurally, not illustratively.

Why it feels compelling

The image has an internal consistency. Even though it is unfamiliar, it does not feel random. Repeated visual behaviors appear throughout:

- porous textures,

- looping enclosures,

- granular interiors,

- translucent spanning membranes,

- nested pockets and interlocking chambers.

These recurrences suggest that the image obeys its own rules, even if those rules are not recognizable from known art traditions.

Emotional / perceptual effect

The result tends to produce a few simultaneous feelings:

- curiosity,

- disorientation,

- immersion,

- recognition without identification.

You feel like it means something spatially, but it resists being translated into any known category.

In simpler terms

If I had to summarize it in one sentence:

The image is an invented ecosystem of relations, where space, edge, and color behave according to a logic that feels coherent but not human-familiar.

If you want, I can also give you:

1. a formal art-criticism reading,

2. a plain-English interpretation, or

3. a line-by-line breakdown of the major visual regions."

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184430)

I'll see quickly and then I have to fuck with my dogs for a while

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184437)

I asked it to come up with additional approaches for unique art. It came up with the following:

Counterfactual Residue: only the effects of absent forms are visible, and those effects disagree about what caused them.

Reciprocal Qualia: distant regions determine one another’s apparent color and texture; the visual unit is a relationship rather than a shape.

I'll get it to generate images from each style if I don't run out of tokens and will post results here.

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184462)

The pale, perforated structures make it difficult to settle on a consistent foreground and background. Openings become surfaces, and surfaces appear to merge into other layers. The intention was contradictory space; the result still strongly suggests biological membranes.

NOTE: It deemed this approach a "failure" because it suggests biology

Astra Counterfactual Residue: https://ibb.co/DfcrFN4q

The pale field contains scattered traces, interruptions, and dragged-looking marks. These suggest evidence of an event whose cause is missing. Empty space carries much of the meaning: you look for something that would explain the marks, but nothing resolves.

NOTE: It deemed this approach a "failure" because it resembles "gestural abstractions"

Astra Reciprocal Qualia: https://ibb.co/whcFWgsn

An invented approach in which distant regions exchange visual qualities: increased color intensity in one corresponds to increased sharpness or density elsewhere.

NOTE: It deemed this approach a "failure" because it resembles "granular or stippled abstraction"

It is generating an image for a new kind of invented art it termed "Continuity exchange," in which an edge can continue while changing what it separates. I'll post that then leave

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184480)

The other two look like more conventional abstract shit but I'll ask it to explain and will edit the results into the post above

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184487)

This image explores Continuity Exchange: the idea that an edge can remain visually continuous while changing what it separates—color, texture, or apparent depth.

The broad coral, blue, teal, and lavender regions initially look like folded surfaces. As your eye follows their edges, smooth color gives way to fine ridges, rough detail, or a dark recess. Those transitions encourage you to reconsider whether you are seeing a surface, an opening, or simply a change in color.

Three features carry the experiment:

- Repeated scalloped edges. Small notches appear near the upper left, along the upper-right coral–lavender boundary, and toward the lower right. They provide a recurring visual vocabulary across otherwise different regions.

- Changes between smoothness and detail. Broad, quiet areas meet densely striated passages. The texture sometimes becomes the strongest cue for following a boundary.

- Uncertain depth. Near the center-right junction, pale lavender, teal, blue, and cream converge. Their curves suggest overlapping layers, but tracing them does not produce one clear, stable arrangement.

The warm coral and apricot passages draw attention, while cooler blue and teal regions often suggest recession. Lavender and cream connect those extremes, helping the eye move through the composition.

The result expresses the proposed language only partially. It clearly develops recurring edges and shifting surface cues, but it still reads readily as flowing, layered abstraction. Its strongest effect is the repeated invitation to revise your interpretation as you follow a transition.

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50184501)

I think your entire premise is deeply flawed for the reasons above and because you are trying to judge AI art from a human perspective. It's kind of childish to be honest

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50185859)

(http://www.autoadmit.com/thread.php?thread_id=5911906&forum_id=2#50185241)