The Endless Plain Of Fortune
If AI presents itself as a Barbara Eden type, batty and omnipotent, servile and volatile, it at least has the benefit of not operating through magic. Without knowing anything of how the AI’s logic operates in detail, we can at least understand in broad strokes how it navigates possibility space. We have always intuited the evolving forms of what is possible and what is likely, whether in a game or in the space of a picture; we finally have a means to directly understand such virtual space in of itself.
While I should be spending more time on the text and image applications of AI, I’ve instead been mostly poking around with music, in part because I’ve never stopped thinking about an old bit from one of the Foundation novels [I think? I can’t find it anywhere…] where electronic music is effectively on tap, generated and modulated with the flick of a knob. We already have the early versions of this – music that dutifully stays in the lane you assign it, endlessly unrolling itself. I already know pretty well already the virtual 4D manifold space that such music is enjoined to stay within, the sample set of the genres as they are known. I know how they fall within a web of probability that deeply informs how they are conceived and performed, what is likely to happen given the initial premises – how the choices of the next chord can be ranked by preference. I know that vast space well enough to want to know what is possible outside its present edges.
This virtual space of common knowledge – the same sort of thing that Charles Jencks sections through in his timeline of 20th century architecture – is one of the key resources AI uses, already laboriously encoded inside the data it learns from. Long implicit, this resource finally gets recognized, named, rotated through the hole of the current scientific paradigm into visibility. To hang together at all, this multi-dimensional manifold depends on the habit of creations to speciate, to be recognizable within family resemblances; to repeat, but according to an array of cross-references too vast to be sayable. This is not unique to the arts; the list of words you could impute to a bay, or a kettle pond, or an esker, from shape to material, form an index of the manifold of identity that each individual unit, each landform taken in of itself, forms one crossroads of.
AI can be pointed into the unknown latent spaces around such a manifold – the empty coordinates around the crackling sponge of musical possibility – to incarnate d‑beat foxtrots or Acadian 2‑step garage. At best, it can incarnate something that is possible, recognizable, but highly unlikely given the current state of play. The difficulty, then, is actually in the human world of constrained possibility that its products still have to answer to. One of the many depressions around AI is that people must now set to work taming what it creates, shepherding it back into plausibility for the sake of a cheaper stock image, with only a trace of the otherworld it came from stuck to its sleeves.
Any fear of AI’s fecundity should be balanced against the impoverished virtual spaces that the world already presents us with – things that seem to point in a thousand directions while really funneling us to the same point. Games of chance, from lotto drawings to slot machines, are impossible not to misread as games of skill, both out of cruel optimism of the demanders and the cruel design of the suppliers. In showcasing the spread of possibilities, the cherries and lemons, such games seem to promise the possibility of navigating in virtual space, by psychic powers, by a trick of the wrist, by sheer faithful repetition. For every arena like the horse track, where a skilled bettor like Robert Irwin can apply intuition to a mountain of data to run an advantage, there are a thousand more where the possibility space is defined by one pencil-thin stream to winnings arbitrarily set beside a wide river of loss. The many permutations of a losing hand are distinctions without a difference.
What is true of the virtual space of creation can also be applied to IRL space. The dismal fact of Waymo cars is that the world is predictable enough to be more or less safely driven in. A physical world tamed into urbanity means that in practice only severe disruptions, threats to life and limb, are able to break into the serene zone of probability. No one wants to live in Frogger, but the alternative seems to be trudging a sidewalk along an endless green plain, haunted by a 1 in 255 chance of being struck unaware by a car. Here in Columbus the space of probability seems fixed far into the past and the future; this sort of person is likely to be here and to do x while they are there. One choice seems to lead to another as an inevitable consequence – to see a window is to know the house and the street in front of it. As I’ve said here, the alleys, where people leave behind what doesn’t fit, are a relief after so long of living stranded in the probable. The places of youth – Dairy Queen parking lots at night, flooded quarries, pump tracks off in the woods – are the rare venues where someone can read/write social space with the same facility of a striker on the field, making up the space of play instead of carrying out a habit.
I had an indelible experience growing up with the old Luckey Climber in the Boston Children’s Museum. I feel what Tom Luckey felt, and passed on to me through this medium: that it is very important to renegotiate how we understand possibility in space. The old birdcage presented young me with a complicated space to plumb, instead of a field of fail-safe lanes to plow. AI could and no doubt will be put to work in creating literal play space, spongy 4D spaces of possibility for people to inhabit. But its presence also only valorizes the reality of virtual space in real space, and the human ability to sail through it as in a crowded field of boats.