Thy Art Has No Soul
Is Art Created by Artificial Intelligence Really Art?
What is Art?
The word art, by definition, involves human creative skill and imagination. Which begs the question: if one removes the human element from what we call art, is it still art?
It is that human creativity, which includes the plethora of things an artist struggles with in shaping their art, that imbues the work with its soul.
What would be the point of looking at a painting that was created by a mindless, non-sentient machine? One that used millions of bits and pieces of scanned art from its database to create a new image composed of a unique combination of that data? (Or, as a few artists have discovered after viewing AI-generated works posted on the internet, a not-so-unique combination, i.e., the work was similar to their own).
Ask an artist to explain their painting: what inspired it, their color choice, its theme, its origin, and they can tell you. Ask a machine those same questions regarding a painting it has created and you will get the AI equivalent of a blank stare — in perpetuity. It will be decades, if ever, before machines can physically create a work of art with the same level of skill and artistry of a human. And yes, I know that somewhere there is likely a robotic arm connected to a computer brain capable of holding a brush and spreading paint on a medium. Even if (or when) it can do so, will it understand brush strokes, shading, blending, texture, or imagery? While it may know the rules of artistic painting via a complex web of algorithms, will it understand which rules it should break in order to give its work a unique perspective?
I’m a fan of the writer Amor Towles’ work. In each of his novels, he uses dozens of words in unique ways. Occasionally, in dialog, he makes up a word. He does so because that is what people sometimes do in real life. And he is not alone in that practice among skilled writers.
Yes, an AI agent can rewrite a passage in a different style and do a credible job. I once asked ChatGPT to reproduce the transcript of a speech given by a well-known politician in the style of Shakespeare, which produced a few hilarious passages. But I note too that starting with the AI’s rework of the speech, I significantly improved its effort, as several of its passages were flat-out wrong, or nonsensical.
Further, ChatGPT’s re-stylization effort did not create brand new passages, new word usages, or new words out of whole cloth, as William might have done. Scholars credit Shakespeare with creating approximately 1,700 new words in the English language, many of which are still in common use today. It was a neat parlor trick, but it was not Shakespeare.
Should AI Create Art?
But whether artificial intelligence can create art masks a more central debate: should it?
Exactly…
Author Joanna Maciejewska asks the important question. Does society want scientists to waste their time and (our) resources (artificial intelligence computing uses an enourmous amount of energy) on AI-driven artistic endeavors? Is creating art the highest and best use of this new technology, especially if society believes that its production is of questionable artistic value?
That view rings even more true if one considers:
AI models that are intended to produce works of art are trained (with a question as to the legality of doing so) using thousands of paintings by true masters: Vermeer, Picasso, Cezanne, Degas, and many, many more. But AI technologists note their machines are fed millions and millions of additional images. You may rest assured that a large percentage of those supplemental images do not rise to the level of masterworks. Therefore, mathematically speaking, only a tiny fraction of the pictorial content the machines have consumed rises to the level of exceptional works. The same would be true with other forms or art created by AI: photographs, books, music, etc. Hence, to Andreessen’s point, on balance, its average in, average out.
Before society travels much further down this road, perhaps it should grapple with the questions and observations being voiced now concerning AI-driven art creation: Is it an endeavor even worth pursuing?
I have learned to never say never within the realm of technology. Never is a long time, and combining human ingenuity with AI and robotic capabilities is a powerful force. It is feasible that someday all forms of artistic creation will be well within the capabilities of a bundle of wires, chips, and composite materials. But even if, at that distant point, AI can indeed create a masterful painting using brushes and oils, what would be the point? The work, just like L. Frank Baum’s Tin Woodman in the Wonderful Wizard of Oz, would still have no soul.
A weekly technology focused newsletter, Launch Key, written and produced by a good friend, Rob Norris sparked the idea for this piece. Check out his work at: Despite the AI narrative, human ingenuity wins.