[ thirty-seven ]


[ recap ]

Last week we looked at Keith Haring's questions about who runs the machine and Milieu, a free record-keeping tool.

[ thoughts ]

Everyone can suddenly make anything, which has a lot of people wondering whether originality is finished too.

But originality was never quite what we think.

Almost nothing we call original was made from nothing. The proof is the invention that made the modern world: the printing press.

Johannes Gutenberg took the screw press that Rhineland farmers used for generations to crush their grapes. Where grapes went, he fed paper and inked metal type.

The type was not his invention either. Korea had been printing with movable metal type for decades, and the way he made each letter came from his training as a goldsmith. The punches used to strike a coin.

None of the parts were his.

What he added was the vision that a wine press and a coin punch could belong in the same machine, producing something new.

Most ideas turn out to work like that.

James Webb Young put it plainly in 1940: An idea is nothing more nor less than a new combination of old elements.

Young's own example for this was a kaleidoscope. Turn it, and loose chips of coloured glass fall into an arrangement you haven't seen before, and the more chips it holds, the more arrangements it can make.

A mind kinda works the same way. The more you put in it (books, arguments, half-heard things), the more ideas it can throw out when something turns it.

A language model turns out combinations too, but it reaches for the likeliest one, the move that fits everything it has already read. That is what it was built to do, and it is why so much of what it makes feels familiar before you have finished reading.

The unlikely pairing is the one it struggles to reach. The wine press and the coin punch, the two things nobody thought to set side by side, is exactly what a machine will miss.

So the improbable connection stays with us.

[ creativity ]

The good news, and there is more of it than the panic admits, is that the machine took the boring half of the job.

What it cannot do is tell whether the thing is any good. Taste is the residue of a particular life, the sum of everything you have loved and thrown away, and the machine has none to draw on. It can make. It cannot judge, and judgement is the aim of the game.

Massimo Vignelli, the designer who gave New York its subway signage, once compared a computer and a pencil:
A pencil does nothing on its own. Leave it on the desk and it sits there, dead, doing nothing until your hand moves it. A computer is the opposite. Even by accident it offers you beautiful things that are very seductive.

So the part that stays yours is the judgement, and it comes from your own life.

A language model has read everything and lived nothing. You have read less and lived more, and the particular set of experiences you carry, the place you grew up and the half-formed thought from this morning's walk, sits in exactly one head.

Those are your chips of glass, and nobody else can turn out the pattern they make, or knows which patterns are worth keeping.

Bring the judgement, and let the machine do the finishing. Noticing is yours to keep.

[ making ]

Young did not stop at the idea. He gave it a method, five steps, and it maps almost exactly what we've previously looked at.

First, gather raw material, the specific kind aimed at whatever you are working on and the general kind aimed at nothing in particular. This is the record from last week.

Second, work it over. Turn the pieces against each other and feel for where they fit. Write down the half-formed ones even when they look like nothing, because most will go nowhere, and that is the job.

Third, drop it. Walk away completely and do something else entirely.
This is the incubation we covered a few weeks back, and Young is firm that it counts as part of the work. The combination forms while your attention is elsewhere.

Fourth, it arrives, usually when you are not looking, in the shower or halfway down the street. Keep the record close enough to catch it before it goes.

Fifth, take it out into daylight. The idea that felt finished in your head comes out rough, so show it, and let people tell you what is missing. The good ones grow once they leave the building.

Point the machine at the first and last step and it'll help. The middle three are yours, and it cannot follow you there. That is where the turning happens, and a model cannot turn what it has never lived.

[ sign off ]

Paul Rand designed the logos you grew up with, and in 1996, before any of this arrived:
"Without aesthetic, the computer is but a mindless speed machine, producing effects without substance, form without relevant content, content without meaningful form."
— From Lascaux to Brooklyn, 1996

Machine speed is real, and I consider it a gift, but aesthetics are still your job.

Stay noticing in the real world.

Thanks for reading. See you next Sunday.

Artificial Milieu

creativity in the age of artificial intelligence

Read more from Artificial Milieu
week [ thirty-eight ], creativity in the age of artificial intelligence.

[ recap ] Last week we looked at originality: Gutenberg's press, James Webb Young on ideas and Massimo Vignelli's pencil. [ thoughts ] In a new city you notice everything: the way doors open, how people queue, what coffee costs, which way to look before crossing. It's not automatic yet, so you notice all of it. Then it fades. Give it a few months and the new street is just the way to the station, the thing you walk through on the phone. I've had that feeling living in three continents, and it...

week [ thirty-six ], creativity in the age of artificial intelligence.

[ recap ] Last week we looked at John Cage's 4'33", Ursula K. Le Guin's Carrier Bag Theory, and Graham Wallas' creative process. [ thoughts ] Keith Haring had just turned twenty when he arrived in New York in 1978. He bussed tables at Danceteria at night and studied semiotics by day at the School of Visual Arts, a year or two from the subway drawings. And he kept journals from his teens until he died. Semiotics is the study of how signs come to mean things and who gets to decide what. It was...

week [ thirty-five ], creativity in the age of artificial intelligence.

[ recap ] Last week we looked at televised essays (Carl Sagan's Cosmos. Kenneth Clark's Civilisation. John Berger's Ways of Seeing) and Michel de Montaigne's thinking. [ thoughts ] I didn't post much this week. By the logic of the algorithm, it means nothing happened. But the feed only knows what has been published. It cannot see the half-finished sentence sitting in a notebook, the image saved without knowing why, pages returned to three times without words added. So it does the simplest...