The Digital Art Conundrum: Is this AI, a Filter or Hand Made?
The question comes up often enough (and I’ve addressed it here and on social media). Images like the one above are a combination of prompt-based AI, neural image style transfer (deep style), and image compositing/collage in Photoshop. I think what surprises me is how little people understand about the process. It starts with Cubism. Cubist collage inspired many artists. It expanded the definition of painting, questioned existing notions of surface and dimensionality, and created a legacy that inspired Surrealism, Dadaism and so on.
Guitar and Sheet Music (see above) was one of Picasso’s earliest collage works. He pasted and layered paper, creating shading and structure. The collages were filled with fragments from popular songs, fake wood-grain paper, and snippets of bodies, faces, instruments and other objects. The sources of these fragments were considered to be found artworks that derive their identity as art from the designation placed upon them by the artist and from the social history that comes with the object.
There is usually some degree of modification of the found object, although not always to the extent that it cannot be recognized, as is the case with ready-mades. Recent critical theory, however, would argue that the mere designation and relocation of any object, ready-mades included, constitutes a modification of the object because it changes our perception of its utility, its lifespan, or its status. — wikipedia
Then, we go from about 1920 to 1985 (or 65 years) when Digital Collage became an art style that uses computer-based tools to make a collages by using, piecing and layering together a great number of images and textures from different materials to create a new picture or piece of art. The first tools I used as a computer graphics student in the 1980s and 90s was Adobe Illustrator (see below) and Photoshop.
One major thing about any collage is that it uses (samples) fragments materials created by someone else… almost always without their permission and with no attribution. Artists could use Illustrator to trace portions of existing images or Photoshop to piece or composite images using found artifacts. After the 1990s, digital collage became an accepted art form and for the next 20 years tools continued to be developed. Then, with the invention of the smartphone, came mobile apps that made it easier for artists and non-artists to access to programs that offered possibilities to make digital art, including digital collage.
With the increased access to technology, and especially the Internet people use the artistic style of digital collage and art for a wide variety of uses. For example, Happy Color (see above) is a “calming paint by numbers game.” When you are done coloring in one of the hundreds of templates available you can share it or save it to your mobile device. Modern digital collage/art became more accessible to the public and relatively easy to do with different programs or tools, but it requires many techniques. I argue that this (modern digital college/art) lead to the development of generative artificial intelligence.
Eight years ago I discovered Deep Dream, a computer vision program created by Google engineer Alexander Mordvintsev that uses a convolutional neural network to find and enhance patterns. The Deep Dream algorithm works by enhancing and modifying images based on the patterns it has learned during its training on large dataset. This development led to “Deep Style” AKA neural image style transfer, which lets users apply style reference images to content or source images. For example, in the image above I used an image I saved from the Happy Color app as the style reference image. It goes back to Cubist collages that modify found material that changes the perception of its use and creates something new in the process.
I used Deep Style for years (and continue to do so) until the emergence of prompt-based AI tools such as Midjourney. It should be noted that today Deep Dream, Deep Style, Midjourney, even Illustrator and Photoshop are all considered to be generative AI. As a digital artist, I often use a combination of tools to create images. I think what people misunderstand is how the current tools, in the hands of experienced digital artists, can be used to create new artworks, with unique visual aesthetics/styles. My intention is go beyond what human hands can do, not make it easier to produce my drawings (see below).
Rather, the purpose of using current tools (including generative AI) is to combine tools and methods to find new discoveries. I can use the image below to address or answer the question: ‘Is this AI, a Filter or Hand Made?’ It’s ALL of those things, all in one image. It makes use of Adobe Illustrator and Photoshop, Midjourney, and Deep Style (Deep Dream Generator). In this new digital space, images (existing works) can be uploaded as prompts with text, or used as style references. To this I add my knowledge of color theory, subject, composition, art history, culture, and style.
In the image above you’ll find references to early 20th century wallpaper and textiles, as well as 1920s fashion to create a portrait of a Black girl. In the prompt are complementary colors such as lemon/orange and indigo, a color between blue and violet. AI tools are trained to do specific things and artists can push the tools to make compelling artwork. Another goal of my AI-generated work is to interrogate historical notions of blackness and gender (ex. girls, women). The prompts I use are a work in progress and the lab with the computer server/bot becomes a digital art studio.
Also, it should be noted that I’ve already been working as a concept artist, dispelling the idea that the tools alone can take the place of artists. It’s a controversial statement to say the least but I’m proof that it is possible.