Logos usually start with an animal, an initial or an abstract shape. The deraison.ai logo started in an algebra book printed in Zurich in 1659, home to a sign you still tap on your phone's calculator: ÷.
Computing did not exist in 1659, let alone artificial intelligence. Mathematicians were already stuck on a question we have never left behind: how do you write down an abstract idea? How do you set it on a page where the eye can find it?
I care about that question more than any other. I have a PhD in mathematics and an artificial intelligence company, deraison.ai. When I redid the site and the logo, I did not want one more stylized brain, one more glowing neural network, one more tech hexagon, because AI logos are already full of them. I wanted to go much further back, to the people who invented the signs we still think with.
1659: an operation becomes a sign
That year the Swiss mathematician Johann Heinrich Rahn published his Teutsche Algebra ("German Algebra") in Zurich. Almost nobody knows the book, which happens to hold a first in the history of mathematical signs: the ÷ of division, printed there for the first time. Rahn also uses the asterisk * for multiplication. To say "therefore", he arranges three dots into small figures, among them ∴.
Rahn's passage on division could hardly be plainer. He announces that ÷ is the main sign for dividing, then says no more. The shape already existed before him as a printer's mark, which typographers call the obelus. In the pages of the Teutsche Algebra it stands for division for the first time.
This story has to be told carefully, the way mathematicians like it told. Rahn was working at the time alongside an English scholar, John Pell, the very man who had taught him mathematics. Who drew the ÷ first, pupil or teacher? Historians cannot say which of the two invented what. We will probably never know.
I like the story better this way. Swiss hand or English hand, three strokes of ink in 1659 turned an operation you ran in your head into a symbol you could see on a page.
A few dots and a line
A division is an idea. ÷ is its visible form. Three millimeters of ink, and a whole operation becomes recognizable, transmissible, usable.
Here is one of the great strengths of mathematics. Besides discovering relationships between things, it invents languages short enough for a human brain to handle. A bar and two dots now do the job of a full sentence.
The deraison.ai logo comes from there. It borrows the grammar of the Teutsche Algebra without copying a single character: a dot, a line, punctuation, separate elements brought into relation. At first glance this small constellation of signs looks like almost nothing. Yet a mathematician distrusts simple shapes, having watched =, +, ∞, ∫, ∴, ÷ go by. Some of the most powerful ideas in our history take up a handful of pixels.
From ÷ to billions of dials
Three and a half centuries later the scale has changed and the problem has not. Artificial intelligence, too, runs on turning the world into computable objects. Take a cat, as an example. The word "cat" becomes a token, a chunk of text. The animal becomes a vector: an ordered list of numbers. Its relationship with the mouse becomes a matrix, a grid of numbers. A photo of that cat becomes plain numbers. A model then learns from all those numbers, tuning billions of internal dials, its parameters. All that tuning yields a representation a machine uses to compute, predict or generate.
Set the ÷ of 1659 beside a model with billions of parameters and the technical gap is dizzying: three strokes of ink on one side, several billion dials on the other. And the work has not changed: we still have to invent a way of writing things down that makes them computable.
I wanted that idea in the deraison.ai identity. Artificial intelligence, in which I see neither magic nor an oracle, continues a very long story: people putting the world into a form so they can compute with it.
Where the name comes from
The name deraison.ai contains a deliberate contradiction. Mathematics is often presented as the land of pure reason: proofs, rigor and axioms, those starting points everyone agrees not to argue about. Artificial intelligence smudges that pretty picture. We build these systems with extremely precise mathematics, then we get behaviors we cannot always explain simply. Everything is computed. Not everything is understood.
That gap gave me the name. In French the word means something like unreason, so it sounds like an attack on reason. I hear it as going beyond what reason explains at first sight, to where you have to try things, measure, understand. Sometimes you even have to let the phenomenon arrive first and wait for the elegant explanation to follow. I enjoy the paradox: we use the most rigorous tools that exist to explore systems that surprise us.
A logo as a program
So I wanted a logo that says very little out loud and a lot under its breath. No robot, no brain, no artificial spark: only elementary shapes. Dots, because before the curve there is data. A line, because between points we look for relationships. Punctuation, because a thought only becomes usable once it is written down. The whole thing nods to the time when mathematics invented its own alphabet.
Look at it: more than a drawing, this logo is a small set of signs. It also says how I approach artificial intelligence: under the apparent complexity of models, go back to the basic structures, look for the right way of writing things down, strip away useless complexity until what is left is simple enough to be understood.
A score, an emotion
Another story of signs has followed me all my life: written music. A staff, black notes, rests, dynamics, ties: to the eye, almost nothing. A musician sees a shiver in them, waiting.
It is the same work as in 1659: catch something invisible, an intention, a feeling, a kind of vertigo, then store it in an object solid enough to cross the centuries. Three hundred years have passed over a Bach score, which has outlived its composer, his instruments, his concert halls. A bow touches a string and the emotion comes back intact, as if no time had passed. The score made the journey for it.
A mathematical proof gives me a cousin of that feeling, in a different key. Music goes straight to emotion, mathematics goes to victory. I know that moment well and go looking for it, a little the way you go looking for the right note: the hypotheses finally lock together and the conclusion cannot be anything else. "It works," you might say. I would call it a small silent triumph. Reality was resisting, then an idea born inside a human head made it give way.
I think this job pleases me for that moment, far more than for a model's score on a leaderboard or the race for more compute: a blurry intuition, still too vague to say out loud, becomes a clean result. For a fraction of a second the whole world fits inside one sign. A well-built artificial intelligence sometimes gives the same sensation: a huge, opaque system suddenly starts doing something right, something sturdy, something useful. Something you can finally show, the way you show a proof, the way you finally play the piece you had been rehearsing in your head.
Symbols outlive machines
Open a 17th-century text today: the sign you recognize straight away gives you a strange feeling. Machines have changed, so have the surfaces we write on, so have the theories. Paper became screens. Printed tables of results were prepared in advance to spare people long calculations by hand. They gave way to graphics processors (GPUs), the chips that models run on. A page of algebra and its handful of symbols gave way to billions of operations per second. Yet like Rahn, Pell and their contemporaries, we are still looking for better languages to think with.
That, perhaps, is what the deraison.ai logo represents: a tribute to an old mathematical story, of course, but above all a conviction about today. Before you can compute something, you have to find a way to write it down. In 1659, that took a few dots and a line. Today, everything still starts with a sign.
Thanks
Thanks to Audrey Dejoux, the artist who designed this website. She put real delicacy into it, as close as she could come to what she understood of my passion for science.
Sources
MacTutor History of Mathematics: Johann Heinrich Rahn
Sotheby's: Teutsche Algebra by Johann Heinrich Rahn, 1659
Wikipedia: Therefore
sign