ABOUT THE PROJECT (Beta / more to come)

For this next step I altered the 3D palm tree models wireframe depictions by applying Gaussian noise until they were barely recognizable and only traces of their former selves remained.

This technique mirrors the foundational process of certain text-to-image AI generators1. Initially, during the forward diffusion process, images are deconstructed with the help of Gaussian noise. Then, in the reverse diffusion process an image is subsequently reconstructed by a model through denoising and digitally manipulating this dissolved image, guided by prompts and probability distribution.

During the encoding phase, an input (such as an image) is passed through the neural network, which transforms it into a latent space representation encapsulating the essential features and structures of the original data (colors, shapes, textures and other visual elements) in a more compact and compressed form. Meaning, that data, in this case, visual information, is obscured from its original form during the forward diffusion process, yet remnants (key features and structures) of its essence persist.

CON[DIF]FUSION (VERSE II) As ink and code converge and blend,
We seek the genuine in a crafted land.
Words once penned by human hands,
Now shared by minds both formed and planned.

In the confusion of data‘s diffusion,
Where every pixel mirrors a grain of sand,
Bits of reality dissolve in confusion
Consumed by the voracious machine‘s command.

In this landscape of innovation and strife,
Where authenticity battles with artificial life,
We tread carefully, seeking the genuine spark,
Navigating the algorithms‘ boundless dark.

The machine perceives and interpretes ‘for’ the human brain which could be seen as what Virilio called the ‘Automation of perception’ and depending on how much has been up to human decisions, also creation.

More information about REVISITED: DOWNLOAD HERE