Mathematician Collapses All Functions to One Weird Formula

追加: 2026-05-05

[00:00:00]plus minus x * y x to the power of y s and cosine. Maths is full with complicated operations that you do to numbers. An ambitious new paper now claims that actually you can do everything with only one operation.

[00:00:16]That's quite a claim. Maybe that finally explains why the Americans have only one math where the Brits have maths. I've had a look. The new papers a preprint that appeared on the archive the other week and comes from a mathematician in Poland. The one operation that he says can give everything is taking the exponential function of a number x minus the logarithm of another number y. He calls that ml short for exponential minus logarithm. For example, if you put x and one into the ml function, you get e to the x minus log of 1. Log of one is zero. So this gives you just the exponential function. You can also get the logarithm itself back out. For this you first take ml of 1 and x. That gives you e minus log x. Then you put that back into the first slot with another one. So ml of ml 1 x1 that gives e to the e overx. And then you put that yet again into the second slot of ml with a one in the first. If you work through the algebra all the extra pieces cancel and you're left with just log of x. So the same function gives you exponentials and logarithms indirectly. But how do we get the standard operations addition, subtraction, multiplication, etc. They're all in there. Take subtraction.

[00:01:35]We just saw that you can express the exponential function and the logarithm through ML operations. If you put those back into the ML function, what do you get? You get X minus Y. You can also get a plus just by entering minus Y. And now you see you can put the sum of two logarithms into the exponent of the exponential function. That'll give you multiplication. And once you have multiplication, you can define powers and so on. If you want to get trigonometric functions, you first need imaginary numbers. You can get those from the logarithm at negative numbers, though there's a branch cut involved there. I know that sounds like a gardening accident, but it just defines the logarithm for a negative number. But once you have that, you get complex numbers and you get s and cosine and even the number pi. What you do not get for all I can see is all real numbers starting from just one and zero, not without taking infinite limits. But the author doesn't make that claim. He says one can get all mathematical operations.

[00:02:43]The paper hasn't yet been peer-reviewed, but I'm reasonably sure it's correct. I saw lots of people online worry about the arbitrary branch cut of the logarithm, but I don't think this matters. Yes, you could choose it differently and that it give you a different definition of complex numbers, but it it still work. Now, you may be wondering whether this is actually useful. The answer is for normal people, probably not. No one's going to replace a normal calculator with one that has a single deranged button that expects you to rebuild cosine by nesting 15 different operations. The interesting part is structural. In computer logic, there's a logical operation called NAND, which is enough on its own to build all the usual logical gates in a computer.

[00:03:30]This paper's trying to find the analog for ordinary continuous maths. The claim is that EML is kind of a NAND gate for scientific calculators and there is a possible practical angle. If every ordinary formula can be rewritten as a tree made from one repeated building block, then that gives you a very uniform way to search for formulas with a computer. The author suggests this could be useful for symbolic regression, which means trying to guess the equation behind a set of data. So this may be less about helping humans do maths and more about giving machines a simpler language in which to search for equations, which is nice because if the machines are going to take over science, then at least they will do it in an orderly manner. I give this paper a zero out of 10 on the meter. It's a cute paper. I think that the point isn't super surprising to anyone, but it is surprising that no one has done it before. There are two ways one can look at this result. One is that mathematics appears more complicated than it is. The other way to look at it is that what we think of as complicated is very subjective. You could rightfully argue that maybe using 10 or so different operations is easier than nesting just one operation 20 times in itself. In the end, it depends on just how you quantify simplicity. This is also an issue that haunts the foundations of physics.

[00:04:59]Because what really do we mean when we say that simple explanations are better if we don't agree on how to measure simplicity? Or maybe it just tells us that mathematicians have too much time.

[00:05:11]A guy just won $1 million by creating an AI film. A 19year-old girl runs an AI agency and earns $100,000 per month. And a recent report from Farber shows that employees who use AI at work are earning 40% more than those who don't. So the question is where are you? Outskill is the first ever AI focused educational platform to accelerate AI learning for people like you and me. They are hosting the 2-day AI mastermind training live this Saturday and Sunday. And right now is the perfect moment to join because you can get in absolutely free for the next 48 hours. Outsklls training has already built more than 10 million AI first professionals worldwide and is rated 4.9 out of five on Trust Pilot.

[00:06:03]When you sign up to outskill, you'll also get access to the 2026 AI survival hackbook, a comprehensive compilation of the upcoming AI shifts in 2026, and the practical steps you can take to be prepared. Seats are limited. Use the link in the description to join. Thanks for watching. See you tomorrow.