RED•EYE WORLD

  • Metaverse
  • Index
  • Team
  • About
  • Aesthetics
  • Beauty
  • Exploring
  • Fashion
  • Gaming
  • Interviews
  • Monday Spotlight
  • Music
  • News
  • Next in
  • Object of Desire
  • Podcast
  • RADAR Newsletter
  • Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst
    Date
    05 AUGUST 2022
    Author
    GLORIA MARIA CAPPELLETTI
    Image by
    GENERATIVE ADVERSARIAL NETWORK ART BY HOLLY HERNDON AND MAT DRYHURST
    Categories
    Aesthetics

    AI and the seduction of the mirror

    Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst
    Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst

    Today’s RADAR by RED-EYE newsletter ventures an analogy between self discovery through AI and the pervasive role of the mirror in human experience.

    Most artists, before the advent of photography, painted their self-portrait with the help of a mirror, which extends our vision capacity beyond normal and is a possible tool to deepen self-knowledge.

    Nowadays images have taken on a life of their own and they resemble aliens memories in constant evolution. We should question these algorithmic visions the same as when we try to understand who we are while gazing at ourselves in a mirror.

    We have placed ourselves in front of mirrors to communicate with gods or unconsciously obtain visual answers to the fundamental question that characterizes humans: “Who am I?”.

    Further questions can be now triggered to AI, as a portal to fantastic dimensions, or cruel reflections of a reality that we do not like. The time has come to remember and accept that there may be other realities other than those that our mind can conceive.

    If you are interested in this kind of aesthetics, and want to know more about how various generative models work, or even just want to keep up with some of the most innovative artists in the field, or if you would like to try your hand at creating some art yourself, make sure you check out the resources below.

    That said, it’s time to try Gaugan, a photorealistic image-generation artificial intelligence demonstration, that allows anyone to generate amazing landscapes using generative adversarial networks or GAN, that was first theorized by renowned computer scientist Ian Goodfellow in 2014.

    Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst
    Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst

    What makes artificial intelligence art revolutionary is the fact that it uses generative adversarial networks (GANs). AI art, or Neural net art, is art created using mathematical algorithms. This AI neural network acts like a computer’s brain, allowing a user to feed data, which has many outcomes, in our case, producing a unique piece of art.

    VQGAN-CLIP uses generative adversarial networks to generate images from text. It is reminiscent of the Arts & Culture application by Google, which promises to map users selfies onto documented artwork. It was built to generate images of fictional celebrities, trained on thousands of photos of existing celebrities. Its creators artificial intelligence used 15,000 existing human-created portraits from different periods in art history as its source material.

    Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst
    Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst

    One of these researchers, Prof. Ahmed Elgammal, has created a system called the Creative Adversarial Network (CAN). Professor Elgammal’s discussions on artificially generated art, and on his system, CAN, a network of AIs that are creative, are especially intriguing.

    Experimental musician Holly Herndon addresses the need for developing new ideological, political, and technological structures and her sophisticated art serves as a rallying cry for novel fantasies and forms of strategic collective action. In her debut NFT collection, Crossing Interface (2021), Herndon along with Mat Dryhurst use machine learning to create nuanced visual compositions that incorporate texts by Iranian philosopher Reza Negarestani.

    In a quest to research possible self reflections and multiple dimensions, Holly Herndon herself and Mat Dryhurst created a set of hand-crafted works called CLASSIFIED using a custom VQGAN as (generator) + OpenAI and CLIP as (perceptor) trained on her appearance and her associated lyrics. According to Holly Herndon herself, CLIP knows who she is, and it has trained the custom network to create vibrant portraits that capture what CLIP knows of her.

    Somehow it feels we are on the edge of a new poetic landscape where AI opens up an experimental technique to mirror the unconscious and explore alternative dimensions.

    Maybe, Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst
    Maybe, Generative Adversarial Network Art by Holly Herndon and Mat Dryhurst

    AI-Generated text edited by Gloria Maria Cappelletti, editor in chief, RED-EYE

    All artworks courtesy of Holly Herndon and Mat Dryhurst

    RADAR by RED-EYE is the first disclosed AI co-generated newsletter, that explores everything we need to know about the future.