Hacker Newsnew | past | comments | ask | show | jobs | submit | fiforpg's commentslogin

The use of computers in mathematics has been somewhat controversial from the very start.

There are of course all the computer-assisted proofs (see 4 color theorem), as well as the partially-assisted ones (see Viazovska et al on packing problems in dimensions 8, 24). But even finding a solution numerically, then rigorously verifying its properties can leave a lingering sense of incompleteness, of a gap in understanding. I like this one quote by (allegedly) Wigner that illustrates it well:

"It is nice to know that the computer understands the problem, but I would like to understand the problem, too."


Reminded me of this quote: the problem with machine learning is that it's the machine that does the learning


A montage is a fantastic device in a movie.

But a montage about weight lifting does not a body builder make.


> but I would like to understand the problem, too

But why should it be the case that this is always possible?

It's entirely reasonable that the set of useful mathematical proofs is a proper superset of human intelligible useful proofs.

In fact, to argue the contrary would imply there is something incredibly remarkable about human cognition.


No, it doesn’t imply that.

Just that the set of proofs a human can interpret and the set of statements a human can understand overlap; conversely, you require that the statements/theorems humans can understand be a larger class than the proofs they can understand.

To me, it’s not obvious which of those should be true:

- can we only understand theorems for which we comprehend their proof?

- or can we understand theorems despite not comprehending the proof structure?

Within the mathematics community, opinions differ. But you’re elevating your perspective on that question into a law, without any evidence.


> understand theorems for which we comprehend

I don't know what your distinction between "understand" and "comprehend" but my point was not about these words, but about being "useful" and being "understandable".

I'm saying there's no relationship between a mathematical statement being useful and it being understandable.

If it is true that "understanding is a prerequisite for usefulness" (where "understanding" means that a statement can be proven in a way that is intelligible to humans) was a property of mathematical expressions, then this fact would certainly be useful (we could exclude any statements that no human understand from the world of useful mathematical expression). But, by that definition, we would need to understand that statement, so you would need to be able to prove that "understanding is a prerequisite for usefulness" in a human intelligible way.

Now what I just wrote is in itself not a proof that we can't know, but proving the above statement would involve expressing the claim in a mathematically verifiable way that was also understandable by humans, which does imply something remarkable about human cognition (something that would be intelligible no less!)


Well, there is something remarkable in human intelligence. We have yet to find anything like it in the known universe. As for the rest, the wise mathematicians are leaning, sorry, hard to lean. TT and co.


But there is. Human cognition is unrivaled.

You're acting as if current-day AI comes even remotely close of surpassing it, lol.

In fact, please show me a useful mathematical proof unintelligible to a human, or even if you would like to "assume" the existence of such, explain how it could be useful to anyone.


> It's entirely reasonable that the set of useful mathematical proofs is a proper superset of human intelligible useful proofs.

If you can't explain something in a way that a child could understands it, you don't fully understand it either.


Mathematics has always been an experimental science to some extent. While Newton, Euler and Gauss would spend a lot time calculating numerical approximations by hand, modern mathematicians have been doing the same using computers and software. And once an a clear picture emerge about what’s going, you can start to formalize that and attempt to prove and communicate your results in the standard definition, proposition, lemma, theorem scheme. (Btw there is even a journal called Experimental Mathematics devoted to this approach).

I don’t see that LLMs will fundamentally change this, but rather accelerate the speed of mathematical research.

Some computer generated proofs might of course be hard to understand, but at least their existence gives another data point work with.

Doing Mathematics is more than proving something, that’s just the end of a long road spent pondering at one’s desk about how things could work out.


Reminds me of a quote from Tsoding

> “Programming is understanding. If you don't understand what you are doing, you are not programming. You are generating text.”

Perhaps a proof without understanding is just generating numbers.


programming is also solving problems

in medicine they use all kinds of drugs which they don't really understand how they work. anesthetics is a great example


By necessity perhaps, but they are desperate to find out why they work so doctors can kill less people. Almost half a million people die each year from medical errors in the USA.

Now they have a valid excuse that the human body is incredibly complex and not yet understood. We don’t have that excuse, because we build all of our software from the ground up. If we don’t understand it, that is our fault.


To bluntly put it in a nutshell, and state the obvious:

If you don’t understand the problem you can’t be sure that the computer does.


Well, if you can formalise the problem statement (this is the hard part) sufficiently well that the computer can produce a proof, you can be very sure the proof is sound.

A fundamental property of any formal proof is that it can be checked by a fairly stupid machine, automatically, because every step is a simple mechanical operation that names one of a handful of axioms and refers to a handful of earlier steps, the truth of which has already been established. So while coming up with a proof may require genius-level thinking, checking an existing fully fleshed out proof is simple -- just potentially very tedious because of the sheer number of steps.

That said, a typical human-written proof omits many steps considered "obvious" to a trained mathematician. Converting this to a formal proof involves interpreting what the original author "must have meant", which requires a lot of expertise and can go wrong -- or it may reveal that there is some inconsistency in the original claim itself.


> checking an existing fully fleshed out proof is simple

The controversy around Mochizuki and the "abc Conjecture" proof is a contrary example.


How does this involve computer checking of a formal proof?

Last time I checked, it was a disagreement over whether an informal proof is sound, which is exactly the problem with informal proofs.

ETA: There might be a misunderstanding about what "formal proof" means. Even a very detailed, precise English-language description of a proof is generally not a formal proof. The bar is essentially: "It could be checked by a machine that follows simple rules." If different interpretations of a "proof" are possible, the "proof" is by definition informal. Informal proofs are valuable because they are strong evidence that there's a corresponding formal proof "underneath" that would establish the theorem's truth, and because they are (usually) much easier to understand.


His partisans are trying to formalize his proof. I expect they're not going to be able to do it, because the proof is flawed.

This is one of the great things about formalization: it would have avoided this entire debacle.


> This is one of the great things about formalization: it would have avoided this entire debacle.

It's also a MASSIVE amount of SUPER TEDIOUS work. And it's the kind of work that folks who think up advanced math proofs tend to loathe. It's along the lines of programming by toggling in the code from front panel switches.

So what current mathematics does is judge a proof by whether or not the application of the proof somehow coincides with the result from some other adjacent branch of mathematical knowledge. So, a "novel" proof is expected to either help prove something in a slightly different branch of mathematics or simplify some already existing proof.

And that is, as I understand it, the crux of the matter with the Mochizuki proof of the "abc Conjecture." Solving the "abc Conjecture" should provide tools for solving other similar problems just like Wiles' proof of Fermat's Last Theorem provided an entire class of tools for dealing with modularity and elliptic curves. And yet Mochizuki seems to unable to do or demonstrate any of that.

So, Mochizuki's work is like someone dropping a gigantic and impenetrable proof of exactly and only Fermat's Last Theorem that doesn't apply to anything else. Sure, it would be an interesting (and true!) thing, but without the ability to use it further, it's a curiosity rather than a pillar.


> It's also a MASSIVE amount of SUPER TEDIOUS work.

It was, but now autoformalization is a thing. If he had delivered his proof and a formalization -- no matter how much it looked like autogenerated slop -- he would have been taken much more seriously.

Despite reluctance to do formalization, I expect it will gradually become required to get math published, since journals will be flooded with AI generated slop. It will be necessary to filter the slop, and requirement of formalization in the supplementary material would be a good way to do that. It would still be necessary to check that the formalization of the statement of the theorem is correct, but that's a much smaller ask.


Complexity theorists are in a good spot


As a programmer I definitely get annoyed when I see code and I don't understand what it does.

But I also definitely don't understand the problem if I can't get the computer to understand it, with tests.

In some sense I always considered programming to be more trustworthy than maths arguments without the certainty of a solver proof.

With all of these questions in the air, epistemology might be making a comeback.


Tests only work for a limited set of programming verification. In many cases you don’t actually know what the output for any given input should be, so there’s no way of verifying the AI-generated code. You just kind of have to trust it. The only exception I can think of is robotics and quantitative trading. Which have already been extensively utilizing AI.


That's a very handwavy way of saying no.

I disagree, software engineering is a mature discipline now, and at this point we have so many testing frameworks (unit testing, syntax testing, regression testing, fuzzing, testing end to end, live, with a subset of known good and incorrect inputs, chaos monkeys, etc, etc, that to say "there's no way of verifying the AI-generated code" is frankly incorrect.

Or, if you insist, defend the "there's no way of verifying the code, at all", and not only AI-generated.

(if it helps I work in the company where before the code even starts being written, several extensive tests for it must be ready first. It's hard to even commit a broken code, and later in the pipeline it's very easy to catch the subtly broken or incorrect code)


  > In some sense I always considered programming to be more trustworthy than maths arguments without the certainty of a solver proof.
But programming is a subset of mathematics. They are both formal languages. I suspect the trustworthiness is more in your comfort level than the ability to verify


That depends on who you ask.

Type theory can also be an independent synthetic foundation atop which you build mathematics.


You can build all of mathematics on type theory? I very much doubt that considering there isn't even a fully unified mathematics. There's holes that don't know how to be bridged between entire subfields. So I'd be impressed if type theory really could do everything, but hey, I don't know


Almost another layer in the peer review process in the best case right? Just a different kind of peer you have to review.


So… more peer review backlog. That sounds fun. Oh, you want someone to review your paper, Mr phd in mathematics with 20 years of experience? Get in line behind chatGPT.


Look up the story of Flyspeck for this taking an entire career.


lean compiles or it doesnt


You can also pass pytest with assert 1 = 1...


There are also ways to cheat like that in Lean, but they are all easily identifiable. So when people talk about formalization, they mean formalization without such cheats.


Are you sure? If an AI would generate a huge Lean proof/program, wouldn't there be a way to hide such cheats in it? Like as in the underhanded C contest?

Because if you give an AI a goal, and cheating at Lean would satisfy that goal, the AI will do it if it can figure it out.


> wouldn't there be a way to hide such cheats in it?

No.


On opening the article, I was somehow expecting a mention of the large deviations formalism, which was (is?) fashionable in late 20th century, and gives a nice information theoretic view of the CLT. Or something like that. There's a ton of deep math there. So having a bio statistician say "look, the CLT is cool" is a bit underwhelming.

Edit: see eg John Baez's write-up What is Entropy? about the entropy maximization principle, where gaussians make an entrance.


> much more performant compared to Linux alternatives is WinMerge

I have found Beyond Compare to be very good on Linux, even on large files/directories.


While you can certainly argue that some texts have more substance to them than these literary works, you cannot deny that most texts have worse prose than the books.


> abusive language/notation basically everywhere in math

In most cases it is not as much abusing notation as overloading it. If you think of the context of a formula (say, adjacent paragraphs) as its implicit arguments (think lambda captures in c++), then it is natural that curly braces can denote both a set and a sequence, depending on this implicit input.

Such context dependent use of symbols is actually rather convenient with a little practice.


"it is natural that curly braces can denote both a set and a sequence, depending on this implicit input."

?

I don't even know where to begin. Overloading symbols in mathematics occurs all over the place. There's nothing wrong with that. The difference between overloading a symbol and abusing it is whether there is an agreed upon definition/convention regarding its use and to what extent its use conforms to that definition/convention. What I'm saying in my original post is that the statement "{x_n} is a sequence" disagrees with the formal idea of what a sequence is and that most writers don't bother to explain their own notational use.

If you wish to re-define the curly braces to have a context-dependent meaning, knock yourself out. But, I would imagine that that practice would confuse a lot of people. Math is a human activity. It's not a programming language.


In the context of double precision the article says

> the largest integer value that can be represented exactly is 2^53

— I am confused as to why it not 2^52, given that there are 52 bits of mantissa, so relative accuracy is 2^-52, which translates to absolute accuracy larger than 1 after 2^52. Compare this to the table there saying "Next value after 1 = 1 + 2^-52".


There's an implied one bit, so you actually have a 53 bit significand (and 53-bit precision) given only a 52 bit mantissa.


Right, I did realize after posting that close to numbers of the form

1{hidden bit} + (1-2^-52){mantissa with all ones}

the relative accuracy — corresponding to the absolute accuracy of a single bit in mantissa — is about 2^-53. The hidden bit is easy to forget about...



While the idea — of shuffling a societal system a little bit to prevent it from going stale — sounds important, I'm not convinced. Random shuffling leads to good results only when it is combined with a good fitness estimate (see: natural selection). And establishing a fitness test for a societal order seems to be a much harder issue than than that of an organized randomization.


Speaking of random shuffling, I think it should be made much easier to conduct RCTs on citizens to try out systems of governance/social programs/etc. to see what works best. Basically test stuff instead of guessing and voting. I think citizens are equal enough if they have equal chance to get to the treatment group.


Much of the fitness test can be from self selection (you apply for a random spot.) Many people wont bother to apply.


Not necessarily—in 3d there are uncountably many non-lattice packings. They all have the same density as the FCC lattice though. To construct these packings, shift horizontal layers of FCC horizontally with respect to each other.

It is conjectured that in higher dimensions, the densest packing is always non-lattice. The rationale being that there is just not enough symmetry in such spaces.


Well these new results (denser packings than before) are regular lattices which might suggest that the optimal packing could be a lattice. (Until the record is broken again by a irregular packing ;-)


That's true — time is a big place, and a lot of things are lost in it. I for one am rather more pained over disappearing of physical objects — genomes, books, art.

Plus, it is all the more exciting to think about what caused some languages to exist and thrive for so long, and the information about the past they retained.


If you like this kind of language archeology, check out David Anthony's The Horse, the Wheel, and Language — for how the people that spoke the Proto Indo-European language were located in time and space.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: