Ai Art as a neurocortical scrub brush

Om G
5 min readJun 24, 2023

--

Part 1 link

As a continuation of trying to stimulate the visual cortex in a relaxing way, I created some more images with MidJourney. The rationale and goals for this project are described in Part 1.

Following are some groups of images created by prompts that were suggested by ChatGPT after some more clarity on the groups of neurons I wanted to stimulate. This project was inspired by the amazing and ingenious Cardano NFT project unsigned_algorithms.

The prompts include shapes specifically, when I was interested in more simple and abstract stimuli, and the symmetry idea is expressed but not in any real novel ways.

MidJourney prompt: Create a series of abstract designs using basic shapes such as squares, circles, and triangles. Experiment with different color gradients to create a sense of depth and distance, and use symmetrical arrangements to create a sense of balance and harmony.

The images produced gained in complexity…

And it even went to a full landscape, which has these calming effects described and for the same reasons, just not as abstract as the task required, which would create a deeper sensation of novelty in the viewer. Being calmed by a sunset isn’t surprising at all.

Then, abstract landscapes with patterns arrived and I was feeling closer to the end product, but not as a direct result of reverse-engineering the visual cortex as I originally intended.

Then I decided to just trust MidJourney:

Prompt: Abstract bold patterns and large gradients meant to directly stimulate the visual cortex at the V1 and V2 neuron groups while relieving effort from the higher cognitive areas.

Prompt image 1
Prompt image 2
Prompt image 3
Prompt image 4

Now the previous images do create sensations but don’t seem to include the specific vertical, horizontal, and angled lines that the neurons are expecting. As proof, here’s MidJourney working on just “abstract bold patterns and large gradients”:

So it seems MidJourney doesn’t “understand” anything about the V1, V2, and V3 areas, and this is not surprising.

I decided to next provide the main topics that visual researchers have discovered about the structure and receptivity of the visual process. I found mathematicians and Nobel prizes that describe neural behavior. I also found work on hallucinations being derived from neural stimulation where a visual experience is reflected from the physical structures.

I included the themes and concepts I discovered as the following prompt: abstract bold patterns with thick shapes, few angles with gradients that repeat simply, stripe hexagon moire, large scale simple patterns, logarithmic spiral, contrast, multidimensional spline, shift-twist transformation creating symmetry

Final 1
Final 2
Final 2a
Final 3
Final 3a
Final 4
Final 5
Final 5a
Final 6

A factor inherent in the unsigned_algorithms that seems to have created the best results here involve rotating the pattern. This is interesting as rotation is a basic element of the unsigned_algorithm equation as well as research by John Zweck in 2003 (pdf).

I find the rotation element very interesting in that other research examines internally-generated V1 activity involving rotation of an image. I can see how this may provide a certain amount of ‘error correction’ when the observed view is internally compared to slightly rotated versions to determine if the stimulus is still translated the same way. With rotation symmetry, many natural forms are still recognizable. Having rotational symmetry in the image may also provide another layer of work reduction for the visual cortex.

The images for me that create the relaxed sensation are eluding to some larger structure like Final 1, Final 2, and Final 6. The intention is a relaxed state and a sense of wanting to view the image again.

Images Final 3 and Final 3a have shapes that I struggle to stop attributing meaning to. I find letters and numbers in Final 3, and the Final 3a image conveys to me wailing of some kind.

Many of the other examples here feel fussy and not ‘subconscious’ enough, i.e. above the level of V1 and V2.

I think I will take a crack at submitting the actual equations derived by Zweck, Hubel, and others, to a large language model in order to create a prompt for use by MidJourney.

I have not yet found any exploration of the psychological effects of viewing things like landscapes and seascapes as they relate to neural structures in the V1, V2 and V3 areas. Most research is directly related to how these basic recognition patterns convert to higher order understanding, rather than what the base level of recognition does to the human mind.

Let me know in the comments which of these images induce a feeling of wanting to come back and view them more, or that provide a sense of deep relief. Thanks for reading :)

Here’s Part III

--

--