There is Nothing Wrong 2019


There is nothing wrong with your television. Do not attempt to adjust the picture. We are now controlling the transmission. We control the horizontal and the vertical. We can glitch, we can burn the picture. We can shape your vision to anything our imagination can conceive. We will control all that you see and hear. You are about to experience the absurdity which reaches from the shallowest inner minds beyond the limits of imagination.

Generated from a machine learning model of 20,000 images of news footage using Art-DCGAN (RobbIe Barrat) & ESRGAN.

Sold as an NFT on SuperRare

digital art with Ai Tools

1024 × 1024 Px 2019

Artist notes:

I used this same training material for Stark Naked Fraud

Anger is the word I started with to put context playing with these images. Is this a fake president or a real one? I really felt like I was living in a reality I could not understand. What was this babbling idiot talking about? Why is he doing these things that make no sense? Everyone around him acts like nothing is wrong but I see the world on fire.

In these art pieces I began with ideas and opinions about deep fakes, propaganda, disinformation and manipulation using powerful and sometimes ham handed technology tricks used to manipulate democracy in the US and the EU. Fake personas were created on social media controlled by a small number of people. The fear of deep fakes was in the news along with AI technology. In addition I had the idea that I was creating fake art. This NFT idea is so crazy some would consider it fake art. On top of that I am using artificial intelligence techniques to generate potentially billions of images by training a model on 10,000 images. So I am playing with multiple layers of meaning around how do we know what is real and fake. Can this art train people to look at things with skepticism?

In addition I am asking: Is this art at all or am I a fake artist with artificially generated images in this new digital NFT space?

Notice how the chryons (crawlers) are picked up by the training. The neural network is trying to replicate the words.

Technical: art-dcgan a project forked from DCGAN.torch by Soumith Chintala  I trained my own model training it by feeding it 20,000 images of trump. I also created additional images using ESRGAN. The original size of the images is 128 x 128 px and the output of the model was the same size. I enlarged it with an online enlargement site.

Collector Bonus Material

faces of fake trump
Rejected images from the Art-dcgan model. A few were turned into NFT art

In the production of art using artificial intelligence or machine learning techniques most of the manual time is spent selecting images to train the model and then time to curate the best output. Sometimes the model needs to be retrained if the output is not what the artist wants.

trump glitched faces from ai model
Rejected AI Glitch Image of Trump
faces of fake trump
Rejected images from the Art-dcgan model. A few were turned into NFT art

From this panel I only selected one or two art works to refine. After selecting one or more image much time spent enhancing, animating and enlarging the images to prepare them for minting. I was excited to see that the crawling text at the bottom started to be legible.

glitched ai images of trumps faces
Rejected AI Glitch Image of Trump

Some Images Used to Train the AI/ML Model

Bård Ionson

11151 Valley View Dr #525 Bristow, VA 20136