This essay changes the focus from the theory-heavy matters of AI and defining art to the more practical matter of the economics of art. This discussion requires making a broad and crude distinction between two classes of art and creators. The first class of art could be called, albeit somewhat badly, “named art.” This sort of art is defined by the fact that its value derives predominantly from the name and fame of its creator. Works by Picasso, van Gogh, Rembrandt and the like obviously fall into this category. Artists who are enjoying a fleeting fame also fall into this category—at least so long as their name is what matters most. This is not to deny that such art can have great and wonderful qualities of its own; but the defining feature is the creator rather than the actual content of the work (broadly construed).
The second class of art could be called, also somewhat badly, “content art.” This is the art whose value is derived predominantly from what it is (which can include the use to which it can be put). Who created the art generally does not matter. For example, a restaurant owner who needs to put up some low-price original art is not buying it because it is a “LaBossiere” (to use a made-up artist) but because she needs something colorful on the walls that is classier than posters and prints. As another example, a starting podcaster who wants a certain music style for her podcasts acquires the music not because it is a piece by the fictional master of music Rock LaBoss but because she needs low-cost music of a certain style. As a third example, an indie game designer who needs illustrations for their adventure is looking for low-cost images that match the style they want and fit the adventure (or at least come within fireball range of those goals). They are not interested in and cannot afford works by some named and famous (“namous”) fantasy illustrator. This essay will be about this second class of art; the first class will get its own essay. Now, on to the second class.
Since the worth of content art is the content, this is the art that is certain to be impacted by AI. Since those purchasing content art are not focused on who created it but with getting the content they want, they will generally not be opposed to purchasing AI created art. This is not to deny that some will refuse to buy AI art for various reasons—such as wanting to support human artists. It is also reasonable to argue that certain content will be beyond the capabilities of AI systems. That said, if the objective of the purchaser is to get a specific content (such as upbeat background music for a podcast or fish themed paintings for a restaurant), then AI created work would be in competition with human created work for their money. This competition would be in the pragmatic rather than theoretical realm: the pragmatic purchaser is not bedeviled about theoretical concerns about the true definition of “art”, they have a content need that they want to address.
Because this would be a pragmatic competition, the main concerns would also be pragmatic—such things as the quality of the work, its relevance to the goal, the time needed to create the work, the cost and so on. As such, if an AI could create works that would match or exceed the quality of the human competition and do so in a timely manner at a competitive price, then the owner of the AI would win the competition. For example, suppose that I am writing a D&D adventure and want to include some original images rather than reusing some stock illustrations that have also been used and reused by other authors. If there was a company whose AI could create such images on demand and for less than what human illustrators would charge, it would be a smart economic choice to go with the AI. As a large-scale example, it would make great business sense for Spotify and Amazon to have quality AI created music for their streaming services—provided that the cost of creating them was cheaper than paying human musicians for comparable quality work. Naturally, they would need to consider such matters as consumer backlash, but such problems could be addressed. Or not, people being people.
While some feel that art creation is something special, this automation of creation would be analogous to incursions of automation in other fields. That is, if a machine can do the job as well (or better) for less cost, then it makes economic sense to replace the human with a machine. This applies whether the human is painting landscapes or making widgets. As with other cases of automation, there would still be a place for some humans. For example, an AI might be guided by a human to create works with far greater efficiency than the works could be created by human artists, but with better quality than works created solely by a machine. While replacing human workers with machines raises various moral concerns, there is nothing new or special morally about replacing human creators—so the usual moral discussions about robots taking jobs apply here. I will not make a special moral case about robots taking art jobs—because there is not one. That said, I will note one distinction and then head back into the pragmatics.
When it comes to art, people do often like the idea of the human touch. That is, they want something individual and hand-crafted. This is distinct from wanting a work by a specific artist—what matters is that a human made it, not that a specific artist made it. I will address wanting works by specific artists in the next essay.
This does occur in other areas—for example, some people prefer hand-made furniture or clothing over the mass-produced stuff. But, as would be expected, it is especially the case in art. This is shown by the fact that people still buy hand-made works over mass-produced prints, statues and such. This is one area in which an AI cannot outcompete a human: an AI cannot, by definition, create human made art (though we should expect AI forgeries of such works). As long as people want human-made works, there will still be an economic niche for the content art created by humans. It is easy to imagine a future in which independent AIs collect such work; perhaps to be ironic. Now, back to the pragmatics.
While considerable work is being put into art-creating AIs, they are still way behind the hype. As such, their invasion of the art-space will be slow one, thus allowing time for the art economy to adapt. Or not, people being people. There is also the possibility that some types or quality levels of art will prove to be beyond the limits of our technology. For example, while it would be relatively easy to make an AI that could write a formulaic buddy-cop script for a movie, it might not be possible to make an AI that could craft a masterpiece of a movie. Then again, it might turn out that the technology can match and exceed all human creations and do so in a cost-effective way. If so, then the creation of art would by humans would be as economically viable as making horse-drawn buggies: a tiny niche. As with other cases of automation, this would be a loss for the creators, but a gain for the consumers. Unless, of course, we lose something intangible yet valuable when we surrender ever more to the machine. In the next essay I will close out 2018 with a look at named art.