Section two

The Machine

920 plates drawn by arithmetic, kept as a control case.

Let me say plainly what you are looking at, because the word gets thrown around until it stops meaning anything. There is no model here. Nothing was trained, nothing scraped, no prompt typed, no one’s hand imitated. Each plate is a rule and a number, a few sines, a handful of subdivisions, and a run of pseudo-randomness that is not even random, since the same number always draws the same picture. The rule is about two hundred lines long, and you are welcome to read every one of them. There is nothing behind it.

That hollowness is the whole point, and it is why they have earned a room. When a picture is impressive you spend your time being impressed, and the question you came in with quietly leaves. These are competent and empty, and because you can see the emptiness you can look straight through it, at the thing that is not there.

What is not there is not skill. The plates are good; a few are better than good, and I have caught myself preferring one to the next, which is the entire argument folded into a small embarrassment. The preference is real, and it is wholly mine. Nothing on the far side of it was reaching for a thing.

The case for the machine runs like this: feed it enough of what people have already made and it will find what those things share, and what they share is the art. The first half is true, and it is exactly why the output convinces. The second half is where it comes quietly apart. What makes a work worth keeping is almost never what it holds in common with the ones beside it. It is the single move one person made that no one else was making, usually for a reason they would explain badly if you asked, and often against every good piece of advice they were given.

This island is full of that move. Adamantios Diamantis gave the better part of a lifetime to walking Cyprus and painting its people, one particular face at a time, and wrote of them as he found them:

the best people I ever knew. Heavy, simple, sure about the good and evil, bearing their long heritage.Adamantios Diamantis, on the villagers he spent decades recording in The World of Cyprus

That is not the average of a face. It is a man who went to a particular village and loved particular people enough to keep them. Stass Paraskos went the other way and paid for it: put on trial in Leeds in 1966, he became the last artist convicted under an English statute from 1838, defended by a row of the century’s finest critics, and he lost, and he carried on, and built a whole college of art out of the wreckage.

Like religion, art is at its best when it’s a simple activity but, as in religion, simplicity can lead to the experience of something incredible.Stass Paraskos (1933–2014), who founded the Cyprus College of Art

A machine is never put on trial. It has nothing it would go to the wall for, nothing it would be ruined and vindicated by. An average cannot manage any of this. It arrives afterwards, by definition, assembled from what already existed and aimed somewhere by whoever is at the keyboard. It cannot want. It cannot tire of its own last five pictures. It cannot be wrong in the way that turns out, years on, to have been the point, and then give nine years to it.

None of this is a prophecy about what these systems will manage. They will get better at the surface, and fast, and a great deal of work that was only ever surface will stop being worth paying for. My claim is smaller, and it is only about direction. A thing whose entire relationship with the world is a heap of what has already been made will always stand one step behind the person making the next thing, because the next thing is not in the heap yet. It arrives second. No amount of scale moves it to first.

The plates are the honest face of the machine: nobody is pretending there is a mind in them. Go and look at the other section afterwards, at Diamantis, at Paraskos, at people who went somewhere and did something no rule could have reached for, and you will not need me to name the difference.

Selected this hour · AC-1260

Model of Two Registrations

flow · Carob · seed 904764167

Streamlines through a fixed vector field. Each of 220 lines starts at a random point and takes 110 steps of five units, turning at every step by the sum of three sinusoids of its current position. A line stops early if it leaves the plate. Roughly one line in seven is drawn in an accent rather than the ink colour, and each is given its own width and opacity. This plate is seed 904764167 in the Carob palette.

Open the plate

A number called mulberry32

The randomness in these plates is not random.

Each one starts with a single 32-bit number and hands it to a function called mulberry32: thirty-odd lines of arithmetic that take a number, stir it, hand back another, then stir that one. The result has the texture of chance. It is not chance and never was. The whole sequence was settled the moment the seed was picked, and it will run the same way in a hundred years, on any machine, for anyone who asks.

So nothing in a plate was ever decided. It was revealed. The line that swerves, the accent landing on the seventh stroke instead of the sixth, the ring that wobbles: all of it was already true before the first pixel, folded up inside a number, waiting to be read out.

The idea becomes a machine that makes the art.Sol LeWitt, Paragraphs on Conceptual Art, Artforum, June 1967

LeWitt got there first, and he was pleased about it. Fix the instructions, hand them to somebody else, let the execution be routine. His wall drawings are still made that way, by other people, from a page of rules. So the question of whether a machine can make art was settled sixty years ago, in the machine’s favour.

But look at what he kept. The machine in that sentence is the idea, and the idea had a person in it: someone who had looked at enough wall and enough art and enough of his own earlier work to know which instruction was worth writing down. The routine part is what he gave away. He never gave away the other part. It was not his to give.

mulberry32 has no other part. It cannot be surprised, because it expects nothing. It has no contact with anything outside its thirty lines: no wall, no weather, no argument with a gallerist, no bad year, no eye that has got sick of its own habits. Its chance is a closed loop. Ours is contact. The dropped brush. The wrong blue that turns out right. The accident that only counts as one because somebody was standing there who cared how it came out. You cannot get that from a seed. A seed has nowhere to be.

Underneath all the scale, that is what today’s machines share with those thirty lines. A model is a far cleverer stir of a far bigger number, drawing on everything already made. It is superb at the routine part. It arrives second, always, because the next thing is not in the pile until a person puts it there.

But they are useless. They can only give you answers.Picasso on computers, Paris Review, 1964. The familiar “computers are useless” is a later paraphrase.

Sixty years on, the answers have got extraordinary. The observation has not moved.

Which leaves a room where a new plate appears fifteen times a day, mostly while the island is asleep, each one perfectly reproducible from a number, none of which cost anybody anything. That is what a picture looks like when nothing was at stake in making it, and there is no shortage of them: the machine will still be drawing at four in the morning, and it will not have got any better.

Everything worth hanging was made by someone who could have got it wrong. Go and look at some. The addresses are in the other room.