Today, OpenAI revealed Sora, the text to video AI which is mindblowingly impressive. The advances in AI art pose the question on how much it makes sense for a human to be in control when making images, videos, or perhaps poems, since ChatGPT is capable of writing those.

A simple way to divide artistic projects is along the aesthetic value axis and the expression axis. The subjectiveness of aesthetics is not relevant here, since the following applies from anybody’s perspective. The AI is optimizing for aesthetic value and it’s capable of producing highly aesthetic pieces in no time. The expression axis, though, is only relevant if you care about AI’s creative decisions. Most people at the time of writing of this article don’t, and conversely, people care about the expression of an artist, and especially of somebody dear to them. I’ve received gifts that were art pieces and the threshold after which quality of such an art piece becomes irrelevant is very low. What’s really the only factor is the expression, thought, care, and the details behind the piece. I cherish those as they’re all associated with the gift giver.

Similarly, when producing an art piece and given the choice to use AI, you have a trade-off. If you want to maximize capturing your artistic vision, you’d want to minimize outsourcing creative decision-making to AI. At the extreme of this, the AI is a small assistant helping your brush strokes, but gets no further say. If you’re writing a poem to express your emotions and qualia, to have the AI convey it fully, you’d need to describe this in immense detail at which the prompt itself is the poem, anyway. The AI might help choose words for aesthetic value or if you can’t think of the right words, but this is a helping hand you don’t need a particularly powerful AI for.

None of this is criticism of AI art. It just points to an axis orthogonal to what AI can optimize for.