Visual Art, Technology, and Self-Expression
When photography was first introduced, many painters dismissed it, arguing 'It cannot be art; the machine does everything for you.' Yet, as time passed, it became evident that human creativity and personal expression could indeed flourish through 'machine production.' Similarly, the emergence of digital photography faced skepticism from traditional photographers who claimed, 'It is not real; there is no film, no tangible substance.' Nevertheless, human creativity found new avenues in this medium as well.
Echoes of these sentiments resurface with the advent of neural network-based image generation: 'It is not real art, a person cannot express themselves with automated image generation using models trained on massive datasets.' To navigate this third historical challenge in the perception of technology within visual arts, let's consider a mental experiment. Imagine generating a random image with neural networks, then, with each click, being presented with four variations, each differing by 30% - a very conservative number, indeed. Selecting your favorite version two dozen times theoretically brings the result 542 times (yes, half a thousand times - check 1.3^24 yourself) closer to your unique personality. This doesn't even consider the impact of textual interaction with AI, reminiscent of the 'in the beginning was the Word' phenomenon. This brings us to the debated 'authorship in the age of generative AI': with a sufficiently long sequence of choices, the result becomes deeply personal. The way social networks monitor your clicks and apparently deduce great depth about you is proof of this principle.
A more robust mathematical illustration of 'deep personalization in the choice chain of image options' emerges when we abandon the nebulous '30% difference' between options. After 24 interaction cycles, choosing from 4 options each time, we access a space of 4^24 = 281,474,976,710,656 possible images. Does selecting one option from 281 trillion potential outcomes (about 35 thousand per inhabitant of this planet) still seem lacking in individualization? However, an even more compelling argument lies in the continual advancement of generative networks and algorithms. As these technologies evolve to more closely mimic human image generalization capabilities and their outputs move away from any resemblance to images used to train the models, the final argument for 'new tool, not a surrogate of creativity' strengthens. Accumulating examples of creations that utilize these capabilities and result from long, individualized interaction chains will further validate this viewpoint.
At every stage of technological intervention in art, there have been those who use it blindly and those who find ways to express themselves. Technology in art is simply a tool, one of ever-increasing complexity. This complexity heightens the focus and dedication required for artists to ride the wild horse of this new technology and reveal their unique vision to the audience, despite the temptation of 'personality-wiping machine self-sufficiency'. Bigger tools will require bigger personalities to curb and pacify them, giving birth to more powerful and deeply personal creations.
Oleg Moskvin, Dec 24, 2023