Week 6 onwards

Looking back at the process.

Besides all the process steps described in the earlier posts, I will try to sum up what changed. I was able to generate a couple hundred interesting ideas every day by continuously running 3 machines with p5js+python+tensorflow+pillow. The codebase has grown to 17 files, which are still in a messy form, and use private services on the cloud for some specifics parts of the process, such as scaling the quality of the generated work.

Ideation became a daily task as I went through Nicolas Bourriaund book, Relational Aesthetics and The Art of Participation: 1950 to Now from SFMoma. I made almost a hundred annotations and highlights throughout the book, which gave me ideas for several works - not only the one I have been pursuing here.

Every day, by exploring the results of the previous one, I was able to understand what could be improved in order to achieve the personal relationship to the work that I was looking for.

Everyday implementation was a coding task that is reflected in my github status:

Feedback was given from everyone I shared the content with. I also explored some physical printings such as the following. This process is being repeated non-stop, creating new works, reading books, getting new ideas, sharing, getting feedback, going back to new work.

Week 5 - Fair Use Doctrine

My work incorporate existing works. There are three types of incorporated (remixed) works:

1. Copyrighted, permission asked: the only content which is visually distinct is the one from my partner, displayed in weeks 3-4. I have the permission to remix it and create the work together.

2. Public domain or CC-0 work. Problem solved.

3. Copyrighted, invisible: a lot of work is used from many authors which provides the basis for part of my algorithms. The result does not copy in any way the features of the original works, besides being transformative in nature:

   1. Purpose: commercial and educational. As we have learned, having a commercial and/or educational purpose does not weight heavily for or against the fair use.

   2. Nature: fictional works.

   3. Amount or substantiality: only the required information is used from the original work, which does not depict the “heart” of the original work. From the title of the original work you can understand the heart of the piece, while my result does not focus on that (the representation) but the computer understanding of that work.

   4. Effect on the potential market of the copyrighted work: the original works are not affected as they present completely different identifiable symbols and representations that are not used at all in my work.

In this last case, my work teaches away from the original works. While the originals are related to either the physical or the digital representations of things and processes, my work represents different aspects of the physical or digital, including critics to unrelated works.

Weeks 3-4.

During those weeks, I have been creating, adapting the process, exploring different inputs, transformations, random parameters and styles. For example, this is the result of a linogravure executed by my partner, a physical representation which has taken random parameters and a chosen style out of 4000 generated ones. There is no photoshop or post processing as in software based post processing.

Ideation Technique. I decided to stick to my current ideation technique, which involves paying attention to everything I get in touch with for a couple of days. My first goal was to invent a magic performance related to something that would appear. As with most goals, it changed.

I was constantly watching artists Casey Reas and Zach Liebermann productions. Zach posts images daily, for years. Casey posts occasionally. Both created the most important creative programming frameworks around: OpenFrameworks and Processing.

I explored OpenFrameworks and Processing code base, looked for its open bug lists in order to try to understand a little bit more about the tools so I could create mine. I found a bug, explored the code, and fixed it. The merge is here.

I understood Processing 4 was the first version to support recent versions of Java. I am a old time Java developer (since approx. 1999) so I could perhaps migrate a library or two. I decided to create a timer library to explore more of processing. I explored the existing libraries and created mine. Here is my first library, already contributed to Processing’s main library repository:

And now it is time to create. Using everything I learned with previous knowledge from AI, TensorFlow and exploring drawings of my partner, here are some of my creations.

I decided to stick to inspirations from 4 artists (one being my partner’s) and using the output practice of 3-4 outputs a day. Here are my latests 4 posts on my instagram. You can find my instagram at @guilhermepessoal.

While Processing and OpenFrameworks have been around for years, the tooling I am developing has been around for one week. It is still closed source as I don’t have the rights to share my partner’s work, but I am allowed to remix her work.

Technique: Based on Neural Style Transfer, with sources from scrapping, glitchy artwork and life. Besides the links in the article so far, an introductory implementation of transfer can be found here.

Evolution: Magic disappeared. I thought I was going for magic, I explored source code, created a library and ended up developing new tooling (so far, for me) and introducing a practice of daily posts of visual artworks. The evolution was described in this article.

Sourcing: all outcomes presented in this article have their sources linked throughout the article. You can follow all links in the article to find them, inspirations, examples, implementations and introductory code. Closed content that is being used in my next steps have not been linked since they will be part of the next steps for the following weeks.