Forgot your password?
typodupeerror
Math

A Mathematical Proof Too Long To Check 189

Posted by Soulskill
from the which-this-margin-is-too-narrow-to-contain dept.
mikejuk writes "Mathematicians have generally gotten over their unease with computer-assisted proofs. But in the case of a new proof from researchers at the University of Liverpool, we may have crossed a line. The proof is currently contained within a 13 GB file — more space than is required to hold the entirety of Wikipedia. Its size makes it unlikely that humans will be able to check and confirm the proof. The theorem that has been proved is in connection with a long running conjecture of Paul Erdos in 1930. Discrepancy theory is about how possible it is to distribute something evenly. It occurs in lots of different forms and even has a connection with cryptography. In 1993 it was proved that an infinite series cannot have a discrepancy of 1 or less. This proved the theorem for C=1. The recent progress, which pushes C up to 2, was made possible by a clever idea of using a SAT solver — a program that finds values that make an expression true. Things went well up to length 1160, which was proved to have discrepancy 2, but at length 1161 the SAT returned the result that there was no assignment. The negative result generated an unsatisfiability certificate: the proof that a sequence of length 1161 has no subsequence with discrepancy 2 requires over 13 gigabytes of data. As the authors of the paper write: '[it]...is probably one of longest proofs of a non-trivial mathematical result ever produced. ... one may have doubts about to which degree this can be accepted as a proof of a mathematical statement.' Does this matter? Probably not — as long as other programs can check the result and the program itself has to be considered part of the proof."
This discussion has been archived. No new comments can be posted.

A Mathematical Proof Too Long To Check

Comments Filter:
  • by EngineeringStudent (3003337) on Tuesday February 18, 2014 @02:50PM (#46278311)

    it is the beginning of AI-science, not the end of human science.

    Science requires testable, provable, repeatable. If a human cannot understand the proof then he cannot participate in the science. This is likely to be referred to as an "early" version of machine-exclusive science.

  • by Kufat (563166) <kufat@3.14kufat.net minus pi> on Tuesday February 18, 2014 @05:38PM (#46279963) Homepage

    Yeah, I'm familiar with SAT solvers and the fact that they aren't REALLY full brute force; I oversimplified it a bit for the Slashdot crowd. Might have gone a little too far on the "lies to children [wikipedia.org]" scale, mea culpa.

    My point was that anyone with high school level math experience can understand the basic problem of boolean satisfiability; I was trying to draw a distinction between problems that are beyond human comprehension and those that are merely beyond human time and ability, with huge SAT instances falling into the latter category. Shouldn't have glossed over the details quite as badly as I did.

  • by Yaztromo (655250) <yaztromo AT mac DOT com> on Tuesday February 18, 2014 @08:03PM (#46281221) Homepage Journal

    SAT is clearly NP complete, and clearly the existence of good SAT solvers is not a proof that P=NP. This means that there will be relatively small problems that SAT solvers won't be able to solve.

    Enjoyed your post, but have to correct a small quibble.

    From a mathematical standpoint at least, being NP complete doesn't imply that there are some problems that are unsolvable; merely that they won't be solvable in any reasonable amount of computing time. If you have a few hundred billion years of compute time available, a SAT solver might be able to solve even those small problems you mention. Of course, from a practical perspective, none of us are going to be here to get the result in those situations, making them unsolvable from a practical standpoint.

    (On the other hand, once the billions of aeons roll by and the machine goes 'ding' and spits out an answer, we do know that we can verify it in poly time. Huzzah!)

    While all of this may seem ultra-pedantic, there is enough confusion about NP out there that someone reading your post may get the idea that things that are NP-complete are unsolvable. They're not unsolvable -- we can typically fashion algorithms to solve them, simply that those algorithms run in nondeterministic polynomial time, and thus may have runtimes exceeding the expected lifetime of the solar system, even with every cycle of compute time ever invented pushed at it.

    ...unless, of course, someone comes up with a proof that P = NP, in which case all those NP-complete problems can be transformed into P problems. Sure, they might still take a few hundred billion years to get a solution, but at least we'd know how many hundreds of billions of years would be needed to get a solution!

    Yaz

We can predict everything, except the future.

Working...