The images and videos that make up the NFTs in this collection, titled Dream JournalML, were created through a very complex process with the goal of capturing images from my dreams. To achieve this, I used cutting-edge neural-network machine learning algorithms in addition to several very involved manual techniques.
First, I compiled entries from my personal dream journal into text files for use with the first machine learning algorithm, which generates images from text; however, it wasn’t as simple as merely pasting the text and then receiving an image. In fact, to get the most accurate visual results, it was extremely important to alter the syntax of the text entries to best fit the design of the algorithm. After this, the images were automatically upscaled by another algorithm. The images generated from this step were interesting enough, but they still weren’t quite right; they needed to somehow be stylized in the form of my subconscious to truly capture the incredibly surreal atmosphere of my dreams. In order to “simulate” my subconscious (abstractly speaking of course), I compiled a set of over 2,000 image files which included original artwork, photographs, paintings, film stills, video game screenshots, anything that I thought I could directly link to the imagery I subjectively experienced in my dreams. This ultimately was the most important step; after compiling this set of images, I used it to “teach” the final neural-network algorithm how to stylize the images generated from the first few steps in the fashion of my real-life dreaming brain. The results are strikingly accurate to the imagery I subjectively experienced, and I consider this artistic experiment a success. Moreover, it has opened up my eyes to the massive potential for more experimental art projects in the future using both neural-network and blockchain technology, and I am now working on a new, even more ambitious project.