Manifesting the Speculative Through Machine Learning
Well, I know that I don’t know nothin’ about what I do. I know that it works. I don’t know how it works, because I know it ain’t supposed to, but it works. I don’t know what it is that I am doing — it works, and to me, that’s the funk. I do the best I can, and I won’t tell anybody that I know what I’m doing. — George Clinton via Vice Magazine
Speculative fiction covers a variety of genres that have elements that do not exist in reality. Coined by Robert Heinlein in 1947, speculative fiction or SF includes supernatural, futuristic, and other imaginative themes and realms. According to Annie Neugebauer, SF texts force readers to imagine (or speculate) on possibilities that do not fit in with their understanding of the world. Here, ‘text’ is used broadly to include any work of art that allows interpretation such as film, television, music, visual art, and design.
According to Marcus Haynes, to speculate on a ‘text’ requires imagining it outside of your frame of reference. For example, what would happen if you asked a machine (computer) to re-create the image of a Husky dog in the style of a turtle? Or the style of many animals? This is what happens when using the original Deep Dream (or DeepDream), created using a deep convolutional network codenamed “Inception” after the film of the same name.
The software is designed to detect patterns in images such as faces, with the aim of automatically classifying them. Once trained, the machine (network) can be programmed to adjust an original image so that the output (e.g. the one for many animals) yields a higher confidence score. After enough re-iterations, the sought features will be adjusted enough that a form of pareidolia results, by which algorithms generate psychedelic and surreal images (ex. DeepDream Husky).
What if we merge the process of DeepDream (Inception) with speculative art or, even better in my case, Black speculative fiction and art? What if the algorithm was changed to generate diverse, representational types of digital imagery? For example, the portrait of Shabazz Palace’s Ishmael Butler (see above) was created using a similar process as DeepDream. However, the result is quite different. There is a surreal-like quality but the image still looks like a traditional portrait.
For a people who have been told constantly that they have no history or future, that they can never be super or a hero, and that their very existence is a nightmare, Black Speculative Fiction allows them to imagine themselves outside of what the world has told them they must be. — Marcus Haynes
I decided long ago that I don’t want to be a traditional or conventional visual artist. However, I want my artworks to convey my experiences. I grew up listening to Parliament/Funkadelic and that greatly influenced the George Clinton portrait (see top image). I saved a specific image style in Deep Dream Generator for the perfect occasion. It was a highly saturated illustration of a cyborg. The result is hyper-realistic and, yet, appropriate considering the subject(s). Initially, I don’t know what the results are going to be but I always have an idea, a speculation about them.
I put this process into use for the upcoming FUTURES exhibition at the Smithsonian Arts and Industries Building. I created eleven portraits of featured futurists, including a self-portrait (see above). I used styles that corresponded with each futurist. For example, I had ‘Earthseed’ in mind when I created the portrait of Octavia Butler. Parable of the Sower is one of my favorites of Butler’s novels and “Earthseed” imagines that the seeds of all life on Earth can be transplanted, and through adaptation will grow. This concept of malleability or change has some connection to the way that machine learning works…
…and I’m still learning and exploring the possibilities.