Play is older than culture, for culture, however inadequately defined, always presupposes human society, and animals have not waited for man to teach them their playing. - Johan Huizinga, Homo Ludens
Last time I checked, if you ask an AI to identify the traffic lights in a set of pictures, it won’t distinguish between a painting of a traffic light and a photo of a traffic light. The thing represented—the traffic light—is the same from an AI’s POV. But to a human being, we might say that the one is a traffic light and the other is an artistic rendering of it. Edge cases might include photos that are aesthetically souped up, adorned, or surrealized and paintings conducted in a hyper-realist style. To Plato, though, all representations, be they photos or paintings, are the same—they’re all not the thing itself: the idea of the traffic light. Thus, to a Platonist, all perception is art in the senses of deception, flattery, and illusion. The AI, trained on a Platonic algorithm would have to say that none of the images are traffic lights—they’re just images.
Even before AI became a trend, we were well acquainted with the idea of a gap between what a test implies is the right answer and what the philosophically pure answer is. Testing well on standardized tests requires not that we tell the test what it ought to test, but that we succumb to its premises. Thus, the meta-test of every test is our tacit knowledge of know to take it and our commitment to doing so. Plato would likely have flunked the SAT verbal.
The simple question of identifying traffic lights is simple from the POV of the goal of the test: prove you are not a bot, but hard from the POV of philosophy. Ironically, philosophers can prove they aren’t bots, but philosophers, only by failing the test. Behavior can never exhaust or demonstrate our interiority. Cue the midwit meme, in which AI and humanity diverge at the edges of the bell-curve and converge in the middle.
For the sake of this essay, I will suggest that art is a type of pretend play, an early sign of advanced intelligence. Play involves the ability to suspend disbelief, which is not the same thing as being taken in. When watching Othello, says Stanley Cavell, you need to believe enough in the play to semi-forget you are sitting in a theatre, but not so much that you become complicit in murder. Thus far, and to my knowledge, an AI can register Desdemona’s murder as either a pretend crime or a real one, but not both at the same time. Of course, in time, it will be able to pattern match art as art, but only by taking a distance from it and seeing it as merely representational. From the phenomenological point of view, art is not simply representation. It moves us because it is a kind of fantasy in which we are allowed to imagine it is the thing itself. Art gives us permission to feel where life does not. (And besides who is to say life itself is not a kind of representation?)
A rabbit instantly knows the fake carrot from the real one. To her, a bright orange toy carrot doesn’t appear as a representation of a carrot, but as a human artifact. It doesn’t even have carrotness. But to the human, the toy carrot is convincing enough that it might be offered to the rabbit as a joke. The joke doesn’t work as well with a plastic burger or a stuffed bunny rabbit or a ball of play-doh or a wooden idol or a copy of the Iliad, for at some point these simply become inedible objects, even forms of cruelty or mockery. The toy carrot’s ontology is context dependent, but crucially it is only a toy carrot for us.
Animals engage in play, but they are their own props. Humans mediate with representations that pretend at not pretending. Realism is the ultimate flex.
To conclude:
1) Art, play, and pretend are signs of sophisticated intelligence demonstrating profound levels of tacit knowledge and theory of mind. The Turing Test for Super-intelligence requires it to be moved in both directions by Othello (“don’t just sit there, do something,” and “don’t get up, dummy, this is theatre.”)
2) Art, play, and pretend confound categorical, binary-thinking, even though they are hyper-logical.
3) Utility makes objects practically recognizable (the carrot as food), but entertainment and art (the carrot as toy) only register for initiates. If rabbits can’t get the joke, how much the more so, machines?
