
The tension and the differences between AI and I as the fuel for creative printed exploration.
The tension and the differences between AI and I as the fuel for creative printed exploration.
The thoughts I went through to achieve this
The tension and the differences between AI and I as the fuel for creative exploration.
All imagery done by me & MidjourneyImage synthesis bridges the designer’s mind and diffused public internet (image) data. Developing an A(eye) is thus exploring and noticing the aesthetic opportunities in the difference between humans and computational semantics.
02.24 “AI eye” synthetic image on how AI is a recontruction of the internet
03.24 Provotype on the “multi-layeredness” of a neural networks
The models behind image synthesis have multimodal capabilities that work within a network of relationships. The model has learnt relevant patterns between image and text and will predict how to recreate images by placing more attention on certain words of the prompt than others, giving them higher weights.
Imagery:
02.24 “Neural Network” synthetic imagery on AI’s semantic relationships
02.24 “Attention” synthetic imagery on AI’s attention algorithms
A denoising model is responsible to generate images that seem to “emerge” from noise. This capability was acquired during training, when the model was fed noisy images and text pairings and learnt how to “clean” these images with machine learning algorithms.
“Denoising” abstract representation of the denoising process
“ 02.24 “Still dumb?” Synthetic imagery on AI’s emergent capabilities
Us humans, we cannot understand the world as diffused, but we can visualise the difference between our reality and AI’s, by analysing the synthetic image. For the AI, there is diffused materiality within patterns images, and thus every prompt can lead onto emergent material qualities that us humans cannot describe (yet). The opposite of symbolic AI; a kind of computational meaning based on learnt patterns akin to
Imagery:
02.24 “Noise cleaner” Synthetic imagery on AI semantics
02.24 “Material Magic” Synthetic image on AI’s materiality
Generative algorithms operate within an aesthetic unity of the training data. This form of orientation within complex networks can be understood as AI “vibes” or “styles” (Grietzer, 2017).
Designing prompts involves is exploring the intersection of design & computational semantics.
Image synthesis invites us into a visual representation of the differences between human and AI. AI has the ability to conjure imagery that transcends the boundaries of natural language but not necessarily of imagination.
The technology anthropologist Bruno Latour criticised how humans approached complex systems such as neural networks without comprehending nor questioning them treating them as “blackboxes”. What if we begun questioning our relationship with the model?
Polo does not judge a designer by his background nor discipline. To Polo you can be an artist, a lighting expert, an interior designer... All at once. In fact, you are just a type of designer, a kind of designer, a style of designer. Euphemisms aside, you are simply an instruction.
Polo places attention to Marco’s prompt and predicts how to turn image noise into Marco’s vision.
Marco removes the noise of his vision and articulates it as well structured prompts
Blackboxing: When a machine runs and one focuses only on its inputs and outputs and not on its internal complexity (Latour, 1963)
a style of designer a type of designer
a kind of designer
The material prompt language system is a system of language constraints I established to treat words as raw materials. I did not prompt around copyrighted material nor brands or cultural styles.
Designing prompts for controlled images involves is exploring
The model turns prompts into smaller pieces of information (tokens). What If I begun noticing the model’s “favourite” tokens and build a vocabulary from there? As a result of using the /shorten feature, I begun prompting in a more concise manner and my image became closer to my visions.
Conceiving prompts as print meant thinking of the print as a layered composition, or an “assemblage”. One which did not need glue or screws, only imagination.
I begun structuring prompts as a sequence of visual layers that interfered with previous layers. I found in this process the ability to design surrealist material stories. For example, one material could be stitched onto the (next) material or weave into the (next) material.
the intersection of design & computational semantics. layer next layer
The Style of You Workshop: I designed a workshop around prompting. The activity asked participants to think of themselves as printed textiles and to describe this textile identity as“vibes and styles”. The aim was to test whether using ontological principles could help unlock participant’s imagination. It also enabled me to get external input to gather more “material” (words) for my exploration of prompting.
Prompting around textile “styles” and “vibes” enabled me to dive deeper into the perceptual qualities of textiles and thus using the model’s computational gestalt. Examples of answers included “ethereal vibes” silk, “decayed style” wool, “pleated style” cotton. This prompting approach also led me to discover “superject” materials.
superject
Borrowed from Alfred Whitehead’s philosophy on emergent entities; to refer to imagery that “emerges”from the combination of the prompter’s imagination and AI.
Mapping the different perceptual qualities between computational semantics and human semantics as a different, more granular and yet wondrous sense of gestalt.
An overview of the layers of possibilities for synthetic image structuring. I used this framework to envision complex images. This method led me to discover that I could add the key word “background” into the prompt structure to orient the model to guid the image in an imaginary Z-axis.
A page on my approach and methodologies for to explore prompt as a craft. This exploration led me to gradually design & refine a prompt structure that I can go back to to create multi-layered compositions.
/blend
Promoting command that combines the aesthetics of multiple images into a new image (Midjoruney, 2023, V6).
motif Motifs are the repeated elements of a design, such as shapes, symbols, or images, that form the pattern
TILE PROMPT STRUCTURE texture + word “background” colours/print colours/print motif* + medium composition
design
design
/ imagine white linoleum background texture, with RGB coloured stripes, a white rope hanging
/ imagine white linoleum texture background, glass beading that looks like cake sprinkles in the movement of a frisbee
/ imagine royal green and bright pink corduroy texture background, divided by a diagonal black rope
/ imagine white, petrol, cream stripes on corduroy texture background, with a vertical nautical rope
PROMPT EXAMPLE
/imagine cotton tetxure background , half ultramarine half navy print, green white and red braided nylon ropes in a diagonal composition
ULTRAMARINE COTTON TEXTILE BACKGROUND
WITH A NAVY PRINT
SOME GREEN, WHITE AND RED BRAIDED ROPES IN A DIAGONAL
Trompe d’oil collection
My final collection celebrates a wondrous look and feel that evidences the AI’s emergent aesthetic capabilities; medium combination, extreme, detailing, extra-ordinary materialities ... I would like to keep these prompts private.
—Trompe d’oil — May 2024
Why concept capture?
AI can be useful to get novel ideas for material possibilities beyond human imagination based on intuitions. The advantage is that one can get engender visual ideas from unresolved mental imagery.
FUTURE MATERIALITIES
CONCEPT CAPTURE
flowerly
OPPORTUNITIES FOR IMAGE SYNTHESIS
/ imagine wooden rock texture background with a white stitching zig-zag moving on top / imagine white creased linoleum texture background, bright red stitching baseball style, moving around like a frisbee /imagine satinated fabric falling with flowery print gradient colours with baseball stitching contours / imagine white creased linoleum texture background, bright red stitching baseball style, moving around like a frisbee V2Synthetic imagery can be easily converted into virtual textures with other AI tools such as adobe substance 3D sampler; which is able to read images, predict and self- render the image’s material qualities.
Image synthesis enables to articulate a sense of reality within the illusion; AI’s stereogram-like compositing responses to text could unlock new possibilities to facilitate the conception of augmented reality experiences for example for game design or AR advertising.
Generative AI could be implemented within user experiences for bespoke printed products (eg bespoke tiles, bespoke textiles). Bots could guide users during generation of design as self- efficient promoting systems.
tile design bot
Hey! I am your future tile. By replying to me, you get to design me. Let’s start?
reply to me with key words ( ) around ...
vibes styles colours prints textures word prompting
test different key words send to production?
Design school could teach how to prompt and think (imaginatively) about design and computational material semantics.
This image depicts a material that gives off __________ vibes
Imagery:
Workshop exercise I did that worked well to train visualisation and prompting. I recommend using an immersive environment like a dark room with a projector to see the student’s proposals!
In pairs your aim is to pick a “vibe” and collaboratively plan-out and test prompts that could visualise it as a material. You will agree on a final image together and note down the final
OPPORTUNITIES FOR IMAGE SYNTHESIS
AI’s ability to mimc styles makes it very easy to decontextualise, depersonalise and disrespect creative identities of brands and independent artists; that’s why I propose prompting around materiality and formal qualities as these words are, in my (A) eyes, new kind of raw materials.
There is no right way to prompt, it is an ongoing discovery so I want to leave you with my strategy but incentive you to explore yours too! The following page is a breakdown of what I found works for me and a template that is versatile for new prompt structures.
GATHERING
PROMPT STRUCTURE
VIBES AND STYLES GATHERING
THANK YOU FOR READING
By Mafalda Reynolds Brandão