Does Art Make the Cut
After my recent post about defensible moats in an AI-powered world, a friend responded to the question of what categories I overlooked, saying
The arts. There is beauty in imperfection. Whether it be singing, art, or storytelling… AI generated art… is soulless.
I have two difficulties with this proposed addition. First, “the arts” are so varied that I don’t believe they form a single category. Second, some of them seem indefensible today, and don’t become more defensible with AI.
The Breadth of The Arts #
Can we treat “the arts” as a single category?
Insofar as performing arts like music, dance, and theater are about human performance, they resemble my Sports category. They are defensible on the basis of their human focus.
Live performance is one thing, but what about recorded performances? These blur the picture. I recently heard a presentation about an AI-powered system to handle customer calls. The company gave the system a new voice based on a six-second recording of someone speaking. What does that mean for voiceover work, book readers, or poetry recitation? How many seconds of video will an actor need to perform before AI can fill in the remainder of their “performance” with generated content?
This use of AI to extend a performance based on existing material looks to me like a variation on mixed-media art. I expect AI to become just another tool for many artists. If so, then “mixed-media” will cover a much broader spectrum of art than before.
What about non-performing arts?
Painting and sculpture involve physical activity and thereby resemble my Physical Work category. But as non-performing arts, the focus falls on the work product and not the work process. In this case I also expect AI to become another tool, for brainstorming, rough cuts, protyping, visualizations, etc.
Current Defensibility of The Arts #
Some people have objected that works of art will retain their value even in an AI-powered world. This conflates the value of the product with the defensibility of the business. A typical artist today struggles to sell enough to support the work. The value placed on works of art are due in part to the fickle tastes of buyers. Particular artists or even forms of art may fall in and out of favor on a whim. Artists have reputedly unstable careers. This does not sound like a defensible business model, even today.
Soullessness? #
One remarkable thing about the broadening adoption of AI is the humanity we read into it. What do Large Language Models (LLMs) do? They produce plausible sequences of tokens. Small wonder that we find them compelling. But we go one step further and read into it some semblence of personality, emotion, even consciousness. Why do we do that?
Have we seen this type and level of output only from other people until now? Is simple persuasion is more of our world than we care to admit? Is our judgement impaired when confronted with norrow virtual communication, without the benefit of body language, touch, and smell?
A sculpture has no soul, yet we see in it the soul of the sculptor. How different is it to see in AI the soul of those who provided its training data? Do we discount that because, unlike the sculptor, the AI is not their intentional creation? Do we discount it as an aggregate of multiple, possibly contradictory, inputs?
Some art may be valued because it was a human production, because of the soul we see in it, placing it in my Status category. Given our propensity to read the humanity in, this strikes me as a weak claim in a weak category.