AI-generated art and music are nothing short of remarkable. With just a few clicks, one can conjure delightful images in mere seconds. For instance, feast your eyes on a photorealistic depiction I generated using ChatGPT, featuring Adam Smith and David Hume sharing a meal, with Hume cheekily leaving Smith to foot the bill:
Or take a look at this whimsical cartoon representation of my summer bowling league team, the Cosmic Colonels:
Both creations came at no cost to me. I simply articulated my vision to GPT, and voilĂ âart at my fingertips.
However, the rise of AI in art presents a contentious debate among artistsâthose who create for passion and those who do it as a profession. The latter group is understandably anxious about the possibility of AI encroaching on their livelihoods. Letâs employ the âeconomic way of thinkingâ to dissect these concerns and forecast the probable outcomes.
We begin with a fundamental principle of economics: demand curves slope downward. This means that as prices rise, the quantity demanded typically decreasesâand vice versa.
AI art is inexpensive for usersâsometimes even freeâwhen compared to hiring professional artists. Thus, according to the law of demand, we can expect an uptick in the consumption of art. This shift indicates that individuals may now seek original art for lower-value applications. The downward slope of demand curves is also accounted for by the law of diminishing marginal utility: as consumption of a good increases, the satisfaction gained from each additional unit tends to decrease. Hence, consumers prioritize fulfilling their most significant needs first before addressing lesser ones. This is why, at higher prices, goods are typically reserved for high-value purposes rather than low-value ones.
As the price of art decreases due to the advent of AI, we see an increase in consumption for lower-value applications. Think about it: creating a cartoon image for a bowling team joke or portraying David Hume as a frugal dining companion. Such uses donât threaten jobs since these images wouldnât have been commissioned at a higher price point; Iâm not hiring an artist for a humorous blog illustration.
Now, what about art that serves higher-value purposes? Could AI potentially replace works that previously catered to these uses, thereby generating increased consumer surplus at the cost of producer surplus? To some extent, yes. Indeed, in a perfectly competitive market where AI art is seen as a complete substitute for human-produced art, we might anticipate total replacement.
However, high-value art tends to serve different objectives. When producing a custom portrait or a commercial product, the stakes are considerably higher than creating a light-hearted cartoon. While AI can replicate certain aspects of these tasks, it remains fundamentally a program, lacking true intelligence. AI does not engage in the nuanced communication that occurs between an artist and a client. Attempting to get AI to produce exactly what you envision can be a perplexing endeavor, often leaving even the AI itself scratching its proverbial head. With human artists, the dialogue is typically much clearer.
As a result, I foresee a market split. One segment will thrive on low-value art, flourishing without eliminating jobs. The other will focus on high-value art, where human creators will likely retain their edge. The critical point of bifurcation will be where job losses occur. Those artists operating on the cuspâthe âintramarginalâ artistsâare the ones most at risk of being displaced by AI. Identifying who these artists are and quantifying their numbers is challenging. Nevertheless, I do not expect AI to pose the existential threat to art creation and the broader art industry that some fear it might.
ââ
[1] A note on terminology: here, âmarginalâ is used in the economic sense. These artists are situated right at the threshold of value, producing work that closely aligns with what AI can deliver. âMarginalâ does not imply a lack of talent; rather, their output is at a value point that is nearly equivalent to AI-generated alternatives.
