Image of Synthesis made by Fine-Grained Text to Image Generation AttnGAN, AI / Machine Learning Algorithm
TEST PATTERN: DANGER is the visual result of the term: Danger, in any declination such as, danger zone, a danger sign, hidden danger, being in danger, dangerous money, in danger of doing and some synonyms such as risk or hazard.
Using machine learning, the machine translates these terms into images based on the training of learning through the network. Sound is also the result of AI processing each image in a soundtrack.
The final result is therefore a video work that represents the image of the concept of Danger created by a machine based on a real training that ultimately translates into “cultural image”.
Text to Image algorithm is made by MIT to translate text into an image.
A single word or a little sentence can be translated into a relative picture
What happens if the Machine tries to depict an abstract idea?
Can a TEXT to IMAGE generative algorithm represent a picture based on a concept?
The result is astonishing!
Merely using a platform about AI and Machine Learning is useful to stress the algorithms and the imaginary that they try to represent. I did lots of trials and the result is an interesting mixing of visual pattern and abstract figures, everything isn’t connected with humans imaginary and that’s the part more exciting!
The pictures produced by the text to image algorithm are related to the “intelligence” of the computational machine and at the moment it cannot represent a realistic storyboard.
Anyway, the result emphasises a such of creative generative aesthetics that cannot slip by to artists and researchers.
The algorithm translation, in this case, is “DANGER” and synonyms into its hypothetical pictures. It’s completely far from the image of the concept that humans can have from their collective imaginary.
That’s the point – the Machine has no idea of the concept that these words represent. The Machine translate words into the picture through the computational eye.
This is the generative aesthetics that represents the media as itself.
CREDITS: AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks
FEMLINK – This videoart work is made for the Femlink “Danger” Topic.
Alessandra Arnò is a multimedia artist since 2000. Her research is currently evolving into the new digital media aesthetic and philosophical scenarios based on media, plus art theories based on the visual culture working with video libraries and collective memories. The other path on which her research is theoretically focused is the representation of the media as “itself” via AI and also ML.
Distributed by VISUALCONTAINER