by Lindsay Curry
An intriguing article was in my Facebook newsfeed a couple of days ago. It was called “Is the Art Market Ready to Embrace Work Made by Artificial Intelligence? Christie’s Will Test the Waters This Fall” (https://news.artnet.com/market/artificial-intelligence-christies-1335170). This got me thinking about the question, can artificial intelligence create art?
Artificial intelligence (AI) can be defined as the broad area of technology that seeks to emulate human skills or human abilities. In the past number of years, there have been many people who have been hard at work trying to produce art through the use of AI. And it seems that Christie’s Auction House is, at least, willing to entertain the idea that what is produced is in fact art.
An overly simplified explanation for how AI creates art is that algorithms, including deep neural networks, are fed tens of thousands of pieces of art as a dataset. The program then produces new works or novel images. Those works or images are either art, or they’re not. Some people say that art is in the eye of the beholder. If that is true, then why couldn’t AI-generated works be called art, so long as it is emotionally charged and inspiring?
It depends on your definition of art, of course. One definition is that art is something that elicits an emotional response; another is that art helps the audience see the world differently. Under both definitions, AI-produced works could be considered art.
Bennat Berger has argued that there are three necessities for something to be considered art:
technical ability, creativity, and intention (https://venturebeat.com/2018/01/30/who-gets-the-credit-when-ai-makes-art/). He states that AI is able to mimic technical ability well and that creativity, while more difficult to code, can be done. “Defined by the use of imagination and originality, it would be easy to rule out (creativity) in AI. But originality can be coded; in fact, machines are unburdened by the limits of logic that keep people from creative breakthroughs.”
That leaves intention. Berger defines this as “what inspired the artistic work, to begin with and what it’s meant to convey, if anything”. But, often people don’t have any background information about the intention of an artist; they judge the art based on how it makes them feel. So is this really a critical part of the definition? Perhaps not.
Some argue that AI is just a tool, like a camera, is for a photographer. In the mid-1800s, many thought that photographs were not real art and that photographers were just machines that produced images. Today, photography is a recognized art form; is AI in the same position that photography was in 1850? (http://time.com/5357221/obvious-artificial-intelligence-art/) Or the same position that filmmaking was in during its infancy when filmmakers were just learning how to use the motion picture camera as a tool to produce art? (http://www.abc.net.au/news/2018-09-01/artificial-intelligence-chris-rodley-on-changing-role-of-artist/10188746)
Technology is simply a tool, goes this line of argument, and artists have been using technology forever. Stone Age artists used flint knives to carve ivory; the Old Masters used cameras obscura to render scenes of great depth. Now AI’s artificial neural networks are the new tools on the block. (https://qz.com/1023493/ai-will-be-the-art-movement-of-the-21st-century/)
Others argue that AI-produced works are the products of collaboration between the human artistic spirit and technology. One cannot collaborate with a tool; one is a partner. In a human-machine feedback loop, a work would be passed back and forth between the human and the AI, each making changes until an original piece is created. But then the question becomes, who is the artist? Who gets the credit, and the proceeds of sale? Is the artist the human who produced the software in the first place? The coder who made it work? The person who pressed “enter” and got the process going? It’s a difficult question, but one that is going to become ever more important legally and financially, as well as philosophically.
The types of work that AI is producing is not limited to visual art; it also produces what are said to be poetry, stories, music, playwriting, and news stories. AI seems to struggle (at least for now) with plot and comedy. But the question of whether AI’s products can be considered art is a question that won’t go away.
There was a study at Rutgers University (https://www.scientificamerican.com/article/is-art-created-by-ai-really-art/) where respondents were asked if four groups of images were human-produced or AI-produced. Surprisingly, a significant number of respondents identified AI-produced images as human-produced. Even more surprisingly, they were asked how intentional, visually structured, communicative, and inspiring the images were – AI-produced images were rated higher than those by “real” artists.
So, if art is defined by the audience, perhaps AI-produced works can be considered art. I tend to think that they can be, but I’m curious to hear your opinion. I am not a visual artist myself, so I may have different conceptions of what visual art includes. I have looked at poetry that is AI-generated, and as a poet, I can see similarities between what is being published and what is being produced. It doesn’t necessarily make it good poetry, but it may still be poetry nevertheless. So the art produced by AI might be art, but not necessarily good art. What are your thoughts on this modern issue?