194 comments

[ 3.5 ms ] story [ 233 ms ] thread
The pattern looks nice, but I would have asked some mathematicians before announcing this as something novel.
I think the invitation is extended by sharing the blog post.
Come on, does every cool thing need a stamp of approval by "experts" before publication?

"SQLite looks nice, but I would have double checked with Oracle engineers before announcing this as something novel"

Well, to some extent, yes.

Otherwise you’d end up like the MD who’ve “rediscovered” numerical integration (the trapezoid method) and got it published in the journal of diabetes or whatever.

So mathematicians should be reviewing papers in the Journal of Diabetes? I think that the Journal of Diabetes should just stay in its lane.
> Otherwise you’d end up like the MD who’ve “rediscovered” numerical integration (the trapezoid method) and got it published in the journal of diabetes or whatever.

This was shocking because calculus is a required subject in American high schools, and this American doctor presumably went to American high school, not because the doctor didn't check in with mathematicians. Frankly, it would be equally shocking if the doctor had asked a mathematician if integration were a thing, because presumably a doctor is an educated member of society.

Nonetheless, even though it's highly embarrassing to the parties involved, with a little bit of self-deprecating humor if I was the doctor, I could tell people at parties that I, along with Newton and Leibnitz have been published on a foundational numerical integration method. ^_^
Calculus is not a requirement to graduate high school in America. Algebra, sure.
It’s surely a general education requirement in most colleges.
It's a blog post, not a peer-reviewed paper. And the tone certainly sounds to me like an invitation to mathematicians "Hey, we're amateurs who found this cool thing, can you tell us more about it?"
I understand what you mean, but how many great breakthroughs in history were where someone shared the foundation of an idea and then another formalized it?
I think it happens fairly often.

In the more mathematical realm in information theory, turbo codes came out of nowhere by people who were not experts in the field of FEC. Their efficiency far surpassed other methods of the time, to the point that their results in the conference paper were doubted. They didn't have any understanding at the time of why they worked well, they just published results.

Lots of physical phenomena start as simple observations that are later worked into theoretical frameworks. The photoelectric effect comes to mind.

>The pattern looks nice, but I would have asked some mathematicians before announcing this as something novel.

The blog post serves that function.

"The initial goal of John Conway was to define an interesting and unpredictable cell automaton....

"While the definitions before Conway's LIFE were proof-oriented, Conway's construction simply aimed at simplicity without a priori aiming at the proof of automaton being alive." https://en.wikipedia.org/wiki/Conway%27s_Game_of_Life

“Fooled by randomness.”
My gut sense is that this has something to do with visualizing the sieve of Eratosthenes, and nothing more than that. But I'd be happy to hear otherwise.
Other than the sieve of Eratosthenes being a way to compute a finite list of primes I don't see the relationship here. Could you elaborate?
I think the relationship is that the sieve has a natural shape since it's based on ruling out multiples.
Isn’t the sieve’s eventual shape just the list of prime numbers?
I don't intent to be rude, but what does that even mean? Can you explain?
Mostly the "pattern holding up until n^2". I don't know exactly how the groups-of-N factors into all this. Maybe it's nothing and there's no clear relationship.
Yep - we are not claiming to be mathematicians, we're pattern hunters... this is an open invitation to others with more expertise to chime in and help figure this out... it's possible it is a minor discovery or even not a discovery... or it could be useful or even very useful. We don't know. Please help us explore it!
Either way, hearing a mathematician explain why this is a non-discovery would still be interesting to pretty much anyone who doesn't study primes. Kudos on a cool visualization.
If nothing else, this depiction of prime numbers looks very satisfying for recreational mathematics, much more so than the Ulam spiral. I think students who learn how to generate this pattern will be inspired to learn more about math.
(comment deleted)
It will be interesting to see if this pattern falls apart when the numbers are reasonably large (like roughly 300 digits+).
”we are not claiming to be mathematicians, we're pattern hunters.”

“Searching for interesting tautologies” or “Hunting for patterns” are good descriptions of what mathematicians do.

Mathematicians do mathematics because they want to be sure that a) they caught a pattern and b) that it is interesting. That’s what’s being discussed here.

Actually there is a pattern. Take a look at the Mathematica code on GitHub.
I think the parent's trying to tell you guys not to sell yourselves short - that what you're doing is exactly what mathematics is.
These visualization techniques have something going for them. I mean, it's amazing work and with new visualizations, people will make new conclusions based thereon much more easily. Analytic continuation ad nauseum. :)
I wonder if there's a pattern to how many primes per cells-with-primes.
Is there are particular reasoning or meaning to each dot being 6 numbers? Is there any significant changes if you pick other numbers per dot, following the same pattern?

Based on the linked explanation here: https://beta.observablehq.com/@montyxcantsin/unwinding-the-u...

I was wondering this myself. I skimmed to the code but can't find any variable for this.
The claim is that it is the same pattern independent of the number of integers per dot.
If the numbers of integers-per-dot was set to 1, the sequence would simply highlight prime numbers as they progress through the pyramid shape since “Parallax Compression” would no longer apply.
The triangle is a hexagon folded onto itself. That's what the "parallax" part is.
(comment deleted)
Yet they say that this only seems valid for the number of rows equal to the number of integers-per-dot that was used. So, it would indeed highlight with reasonable certainty all the primes in the first row. (I am not a mathematician but I think that is "1" ?)
Because primes exist on a base 6 numeric scale. Where all prime numbers are multiples of 1 and 5.
... I think that statement is only true for the prime number 5. ;)
Base 6:

1,2,3,4,5,6

7,8,9,10,11,12

13,14,15,16,17,18

19,20,21,22,23,24

25,26,27,28,29,30

See a pattern? 2 & 3 are the only prime numbers that are an exception. All others are multiples of 1 and 5.

> All others are multiples of 1 and 5.

Assuming you used “and” when you meant “or”, that's trivially true (and redundant) in that all numbers (irrespective of base, which has no effect on this) are integer multiples of 1.

But no primes other than 5 are integer multiples of 5, in any base.

> Senary may be considered interesting in the study of prime numbers, since all primes other than 2 and 3, when expressed in senary, have 1 or 5 as the final digit.

https://en.wikipedia.org/wiki/Senary

I expressed it poorly, but not as you state, incorrectly.

> I expressed it poorly, but not as you state, incorrectly.

No, really, it is completely incorrect to use “multiple of X in base Y” to mean “have X as the final digit in base Y” (which is equivalent to “is congruent to X modulo Y.”)

13 is not, in base 10 (or anywhere else), a multiple of 3.[0]

That's just not what “multiple” means.

[0] Well, the number denoted by the digits “13” in any base that is itself a multiple of 3—other than base 3 itself where “13” is not a valid number—is a multiple of 3, obviously, but we're talking about the number represented by “13” in base 10.

Having x as a final digit does not make a number a multiple of x. For example, 13 is not a multiple of 3 in base 10.
> Because primes exist on a base 6 numeric scale. Where all prime numbers are multiples of 1 and 5.

What you seem to mean to claim is that primes > 3 are all congruent to either 1 or 5 modulo 6.

From a cryptography standpoint, could this hint at attack vectors for things like discrete log problems? I've only learned of the math behind that myself recently, not sure what implications having a "topographical map of the primes" could have, especially if the pattern is relf-repeating regardless of the size of the primes.
Yes, but only up to a certain point. Larger semiprimes are still a pita to get the primes out of.
> on January 18, 2018, I found a numerical sequence that generated the exact same pattern as Shaun’s pattern

Does this mean that we have a sort of bloom filter-esque test for primality? (ie, it will give you a guaranteed no in O(1) but you'll have to crunch numbers to get the yes?)

If so, are there implications for things that want to know "is it prime?" quickly? Crpytography comes to mind, for instance...

>Does this mean that we have a sort of bloom filter-esque test for primality?

This catches most non-primes ;)

bool maybe_prime(x) { return x % 2 && x % 3 && x % 5 && x % 7; }

And it's infinitely expandable to improve accuracy, at a linear cost of computation!
We already have quick algorithms that say "is it prime" with certainty. Reducing the required time from O(log^6(n)) to O(1) isn't particularly important from cryptographic point of view.

https://en.wikipedia.org/wiki/Primality_test#Fast_determinis...

Note that log(n) is the length of the prime, so if you are using 2048 bit primes, log⁶(n) is quite large. I don't think anyone actually uses one of the general deterministic primality testing algorithms in cryptographic applications.
It also doesn't reduce the time to O(1). Each of the ranges is of size O(2^(N/2)) for an N bit prime, so it's really not useful at all.
we do not have a fast, certain, algorithm for large numbers. The only thing we have for large numbers is “it’s probably a prime. if it’s of a special form, such as messenne, use this other algorithm that takes like 10 days to confirm whether it is actually a prime. If it’s not a number of a special form for which we have specialized algorithms for, which are still pretty slow for large numbers, well you’re SOL, and if you really want to know for certain if it’s a prime or not you better use one of the clever brute force algorithms that do things like only check odds and only check up until the square root of a probable prime, and are still slow as all get out”

So speed improvements are welcome for academic use. Although this doesn’t look like it’s a game changer.

If by "bloom filter-esque", you mean a probabilistic way of testing whether a number is prime or not, then the answer would yes for a majority of numbers.

How? Just run Fermat's little theorem on multiple values of "a" until you feel comfortable. [1]

Why a majority? There are certain exceptions such as the Carmichael numbers to which we need to use slower algorithms to verify the primality of.

Note that I'm assuming the number of calls to an algorithm is constant since it would be naive to discount the size of the number you're testing.

What are the implications of this result? Off the top of my head I can only think of one: the problem of deciding whether a number is prime or not is in the complexity class P (decidable in polynomial time). [2]

How does this affect cryptography? I would say not by that much.

Why? The "hardness" of some forms of cryptography (asymmetric) isn't from the ability to determine primality of a number. It's from the ability of determining the factors of a number. [3] That problem itself has greater implications for the field.

The question of whether it is possible to do so in polynomial time on a classical computer is actually an open question in CS right now.

Note the emphasis on classical! Amazingly, there exist a polynomial time algorithm to do so on a quantum computer called Shor's algorithm. [4]

[1] https://en.wikipedia.org/wiki/Fermat_primality_test

[2] https://en.wikipedia.org/wiki/AKS_primality_test

[3] https://en.wikipedia.org/wiki/Integer_factorization

[4] https://en.wikipedia.org/wiki/Shor%27s_algorithm

Fast primality testing is of course important from asymmetric crypto. Without it, we couldn't generate big semiprimes.
Just run a search on: prime numbers fractal, and many interesting articles (both academic and other) will surface. Search for: zeta function fractal, and you'll find more. I'm no mathematician but it wouldn't surprise me if two very deep fields turn out to be different "manifestations" of some deeper underlying truth, and I then wonder about the possibility of a larger grand unified theory that includes both maths and physics that would have far reaching insights which would go beyond both mathematics and physics, but then I realise that perhaps I've seen too many movies :-)
If this can really map prime numbers to a least a general region, would we be able to break Diffie-hellman key exchange more quickly?

If so this could be a huge blow to security.

But I must say this is amazing that they were able to visualize prime numbers in this way. These guys are geniuses.

I was initially thinking the same, but I don't think that's the case. This will help you find a small range of six potential primes (pick any element on the left edge of the triangle). But you're still required to calculate the factorizations on those six numbers to determine which of them are indeed prime.
No, this won't help with finding primes. As they noted, the pattern for ranges of size k only holds for k lines. So to find a prime of length N (on the order of 2^N), we need to have k=O(2^(N/2)). However, this yields a guarantee that a prime lies in a range of size O(2^(N/2)), which is not particularly useful. We get exponentially better probabilistic results from the prime number theorem.
Just a heads up, but I would pay money for mugs/t-shirts with the pattern. Of course I can just use to code to make it myself, but it's nice to support people sometimes
A few quick observations:

- The right edge of the triangle is always red (ie, no primes present), because it represents (row number) * (3* (row number)+[1..6])

- Prime numbered rows are always black (except the far right column). I can sort of feel why this is true but can't express it mathematically yet.

It's equivalent to http://oeis.org/A054521

Cell i of row j contains n elements of an arithmetic progression, with common difference of j. If i and j have a GCD != 1, they are not coprime, so they share a factor p, as do all numbers in the sequence, so they cannot be prime

Otherwise it's very likely there's at least one prime

If j is prime every column i is coprime with it, so it's going to be black

The Telegram link leads to a distribution channel, not a group, so one cannot actually discuss it there. I let the author know through the contact feature on the website, as it might be unintentional.

Edit: The link was changed to a group (at the bottom of the post).

Awesome! It's the math enthusiast's dream to come up with something new and exciting outside of academia. Recently I discovered what I thought was an interesting chaotic map, but after posting a question about it to Math StackExchange[1] and emailing one or two professors (no response, which is understandable), my obsession waned and I gave up on trying to figure out if it had any significance. Maybe I should keep trying!

[1] https://math.stackexchange.com/questions/2654984/identifying...

Keep going!

Don't forget: There's nothing bad about no-response - People are much more likely to respond on the web and email when you're wrong. ;)

> There's nothing bad about no-response - People are much more likely to respond on the web and email when you're wrong. ;)

There certainly could be something bad about no-response. As someone in academia who gets uninformed musings or crackpot theories from laypeople in his mailbox from time to time, no-response basically means “I know you are wrong, seriously wrong (and, in many crackpot cases, probably mentally ill), but I am not going to just waste my time trying to tell you that.”

As someone else in academia who gets his share of crackpot emails, no response can also mean "oh god, I don't have time to think about this when I have 1000 other emails from my students and colleagues".

(I'm particularly horrid at email.). So I wouldn't take it personally.

It sounds to me like you should have a prewritten response for people who were interested enough to contact you, but who don't know where to start.
I am sceptical of the ability of laymen to even meaningfully understand the field that I am in enough to contribute their own ideas. The amount of literature that would have to be read and assimilated is vast, really achievable only for academics. My prewritten response would only be “Do at least an MA in this field, then we’ll talk”.
I genuinely think that's a good response.

Maybe the majority of recipients won't act, or they'll even take offense, but if somebody in there chose to contact you because they look up to your position, you could easily find in a few years that you challenged somebody to get on the right track with one of those responses. Some people are on the fence and looking for a push.

Obviously it's up to you, but I think it's worth a shot.

These runes would serve excellently as a written alphabet for an alien species in fiction (or maybe for modrons or other Lawful creatures in D&D).
I'd like to see the triangle fractal as a starting state for Conways "Game of Life".
As displayed in the image, every other row is offset by half a square from being on a grid.

Maybe if you shifted everything to the left to line it up, but

Who said you couldn't have a hexagonal lattice? Instead of 8, every field has 6 neighbors. The only problem with that really is that you have to expect the 23/3 rule to look differently on that grid. So, no standard glider.
(comment deleted)
"It looks like something they found on the ship at Roswell."
Is each row the same as if the remainder on division of the cell number by the row number is 0 then red, otherwise black?
(comment deleted)
I'm noticing a few neat things about there is left-right reflected symmetry.

Really neat! (I still know I haven't the math chops though. Good place to further my study)

(comment deleted)
My take, but correct me if I'm wrong.

Take the relation to the GCD triangle http://oeis.org/A054521

At GCD(n,k)!=1 all numbers are divisible by GCD(n,k) therefore contains no prime

At GCD(n,k)==1 we have https://en.wikipedia.org/wiki/Dirichlet%27s_theorem_on_arith... - so those series contain infinitely many primes - and seems like they actually contain at least one prime in all the pixels of the first N rows (but this should be explained/proved, if it is always true for any chosen N, or just happen to be true for the N-s tried by the OP)

You're onto it: what you say would still need to be proven is actually not even true, see anderskaseorg's comments, e.g. https://news.ycombinator.com/item?id=17104624.

So their picture is nice, and the gcd!=1 cells are all red, but the gcd=1 cells need not be black. For odd N this fails loads of times (always?), and for even N you quickly find failing cells when you start looking for it (see linked comment).

(comment deleted)
Does this get us closer to being able to generate the Nth prime without factorization / primality test of every number?
No, very far from it