How can intent be communicated when ideas are filtered through a pseudorandom number generator? Simply looking at the output does not make it clear whether a stroke is placed on the canvas because of a decision by the artist or by a roll of the dice. The artist’s hand is more present in the process than the result; they choose the algorithms, tune the parameters, and prune the features. As I experimented with my code, I held a particular idea in mind, but I was uncertain how it would finally manifest. I pursued a visual composition that made me feel a certain way, and the visuals I landed on communicate my feeling better than anything I can put into words. Through Curved Air is about the challenge of communication. Regardless of the medium, I can never be certain that the message I intend to send is the message that is received.

Through Curved Air takes inspiration from musical minimalism, where repetition is used to highlight changes in rhythm and harmony. Every output is created by repeatedly copying a single shape. Each copy only has meaning in relation to the other shapes nearby, and that meaning changes with the context around it. Delicate structures form and are subsequently obscured by further layers, and though they may be hidden, they still affect the perception of the final image. Clicking on the canvas will start or stop the looping mode, where the image is recreated in a slightly different way. New facets are exposed as the composition changes yet remains the same. Just as I spend my days trying to share my inner experience through my art, Through Curved Air uses each shape and each loop and each token hash to attempt to create a common feeling between myself and the viewer.

Through Curved Air also takes inspiration from the paintings of Joan Mitchell. I was drawn to her expressive strokes, vibrant colors, and use of negative space, all of which come together to give a more intense feeling of looking at a landscape than just looking at a painting of a landscape. Her energetic marks clearly communicate that there is a human being behind the art. The marks of generative works are often similarly representative of the underlying process, but instead of drawing focus to the artist behind the piece, they draw attention to the algorithm. To convey human emotion in the way I intended, I had to use a subtler algorithm.

The technical tools to convey this emotion are tools I have been consistently developing in my practice. I love making generative art because I can design a system where I understand how all the components work, but they can still come together to produce a result that surprises me. The algorithm behind the compositional variation in my earlier works Coquina and Primordial is one I continued to evolve, and it rests at the heart of Through Curved Air. In fact, I pushed it further than ever, using it to create the individual shapes that form the building block of each output. This emergent variation permeates the work and results in surprising images that more closely resemble paintings than anything I made previously. I believe the potential for surprise is key to conveying emotion through code. When the possibility space of the algorithm grows too large to understand, the mind wanders to the person behind the algorithm. Perhaps the human element is the struggle to understand, or the struggle to be understood.

Through Curved Air

Aragnation questions preconceived notions of what AI art can be. Does AI art have to be powered by giant neural networks? Do the aesthetics have to be that of photos or paintings or collages of pieces thereof?

Aragnation uses tiny AI models and constructs abstract imagery in the vocabulary of abstract blobs and organic textures rather than pixels. It is playful with colors. It has a hint of a watercolor and art paper aesthetic. 

The training data includes photographs taken by one of the artists, Abhishek.  The generated pieces thus capture, in broad strokes, Abhishek’s photography – nature, landscapes and urban settings.

In a refreshing contrast to the gravity of the role AI technology is playing in content creation and creative expression, Aragnation brings a light and joyful touch to generative AI art. 

No GPUs were harmed in the making of Aragnation. All AI models are entirely different architectures than those commonly used, and have been designed and trained from scratch by the artists.

Aragnation

“Phototaxis” is a simulated environment populated by simple machines. After Valentino Braitenberg’s 1984 book “Vehicles,” there are four kinds of machines with different wiring and behaviors: Explorers, Aggressors, Lovers, and Cowards. Each line charts the history of one machine’s movement. The core of this code was the foundation for my MicroImage series of works dating back to 2001. I’ve worked on it in fits and starts since then, and in 2021 it was further developed for Art Blocks. Press ‘P’ to play/pause, ‘B’ for the big bang, ‘1’–’5’ to change the magnification, and ‘L’ to reveal the lights’ locations.

Phototaxis

## SL/CE (slice)
This artwork explores Suprematism and generative composition, attempts to fill some gap between generated and constructed artworks.
It achieves composition by slicing simple shapes into pieces by some rules and testing them against a "bad composition detector", then carefully selecting colors for shapes.

### Algorithm
Each iteration starts with selecting the simplest figure (circle, square, rectangles, trapezoid), then shape is placed on the canvas according to specific rules, cut along one or more "slice lines", after which the parts can be shifted, rotated or removed. 
The process is repeated several times for several starting figures, after which the resulting composition is obtained. 
Then an "Bad composition detector" function comes into play, and if the composition is considered "bad", the pseudo-random composition process is repeated with partially different inputs until a "good composition" is obtained, according to this evaluation function.

### Bad composition detector
Function that takes into account and evaluates the size of the figures individually, the total area occupied by the figures, distribution of the figures on the canvas, the balance of emptiness and fullness.
All of these parameters are used to find those outcomes that seem to me to work poorly. It turns out there is a piece of me in this code.

### Result
Looking at images, you can often see something in them, this effect is close to pareidolia, sometimes images works well as profiles pictures, and I am very pleased that I was able to achieve this.

### Extra
Each token is a unique generator with partially fixed parameters.
Click or tap at the bottom half of image to generate a new image.
Click or tap at the top half of image to toggle slideshow.

SL/CE

Paths is an animated journey across various subterranean and aboveground environments. To fully experience the project, you need to see it in high resolution with a capable GPU. Make sure that your browser uses your dedicated graphics card if any, and that you get a good frame rate. Control keys: 1-4 (change speed), 9 (fill aspect ratio), 0 (original aspect ratio), space-bar (play/pause)

Art Blocks Presents

Paths #424

Darien Brito

More in Art Blocks Presents

Through Curved Air

Jacob Gold

Aragnation

Devi Parikh and Abhishek Das

Phototaxis

Casey REAS

SL/CE

Stranger in the Q

Art

Company

Learn