For example,how about in this mathematics if If we calculate to the 80th power of 10 and still don't have a single counterexample, we can say that we have "proved" Goldbach's Conjecture, because the number of atoms in the universe is no more than 80th power of 10.

"But consider: if our immersion in time is taken to place severe constraints on these, then why not similarly our physical limitations? Afterall, there is a perfectly good sense of ‘could’ in which none of us couldconstruct, or survey, a finite segment of that was so big that it included more members than the number of atoms in the known universe, or more members than the number of milli-seconds that will have elapsed by thetime the earth has been swallowed up by the sun.So cannot these arguments casting doubt on the idea of an infinite co-incidence be extended to cast analogous doubt on the idea of a truth concerning some sufficiently large natural number?Wittgenstein may have thought that they could.He sometimes vergedon a correspondingly extreme position.28 Others have recently explored the position more or less sympathetically.29 But it finds no place in either Brouwer or Dummett.Indeed Dummett has argued that it is incoherent(there being no coherent way of saying what is meant by ‘sufficiently large’)"(A.W.Moore, Infinity, second edition, Oxford)

  • By the way, Feng Ye's Strict Finitism and the Logic of Mathematical Applications deserves to be mentioned. Jan 19, 2021 at 13:00

1 Answer 1


The key term is "ultrafinitism."

In one sense it's not hard to whip up such a system. Here's a straightforward, if naive, development of a semantics for arithmetic which basically follows your first sentence:

We pick some number, say x=10^80, to be our limit on the "actual natural numbers." With x in hand, given a sentence p in the language of arithmetic we re-interpret p by treating the quantifiers as ranging over 10^80 as opposed to over all natural numbers.

For example, the standard formalization of Goldbach's conjecture is basically as the following sentence (X):

  • For all x>2, IF there is some y such that x=y+y, THEN there are a,b such that

    • a+b=x,

    • for all p,q, if p*q=a then p=a or q=a, and

    • for all p,q, if p*q=b then p=b or q=b.

Our new semantics is defined via a syntactic translation: to tell whether (X) is true in our new semantics, we ask whether the related sentence (Y) is true in our original "infinitary" semantics:

  • For all 2<x<10^80, IF there is some y<10^80 such that x=y+y, THEN there are a,b<10^80 such that

    • a+b=x,

    • for all p,q<10^80, if p*q=a then p=a or q=a, and

    • for all p,q<10^80, if p*q=b then p=b or q=b.

Note that this isn't exactly what you might expect (namely, "There are no Goldbach counterexamples <10^80"). However, this discrepancy vanishes under further examination since in the sense of our old semantics, the above (Y) is itself equivalent to the following sentence (Z):

  • For all 2<x<10^80, IF there is some y such that x=y+y, THEN there are a,b such that

    • a+b=x,

    • for all p,q, if p*q=a then p=a or q=a, and

    • for all p,q, if p*q=b then p=b or q=b,

the point being that all but one of the "boundings" involved in the translation above were actually redundant. So new-(X) is old-(Y) is old-(Z), or more clearly we as expected get:

Goldbach's conjecture (this is (X)) is true according to our new semantics with x=10^80 iff there is no counterexample to Goldbach below 10^80 (this is (Z)) in the sense of our old semantics.

This may look really really really pedantic at first, and it is, but it's important to demonstrate that this translation is in fact totally precise and something we can reason with.

(Incidentally, this general type of translation is so important in mathematical logic that it has a name - we generally talk about relativizing a formula to an appropriate fragment of a given structure.)

However, there are real problems here. First, not all math is directly about the natural numbers. How do we bring (say) measure theory into this context? Second, the choice of "bounding number" x was pretty arbitrary, yet obviously affects the resulting theory we get. But let's say we only care about basic number theory and we don't mind arbitrariness. Then we still face the ontological issue: our new semantics was defined using our old semantics (this was the (X)->(Y) translation above). Really it seems only satisfying if our new semantics is itself "developed below 10^80" in some sense. However, here we run into a complexity obstacle: very roughly, it takes more than k-many bits of information to describe the first k-many natural numbers. So while we can produce something which looks like what you're describing at first glance, it's not at all clear that anything satisfying has actually been done by doing so.

Of course this was only a very naive first step, and there are much more valuable approaches to take (see e.g. here or here). However, it is nevertheless the case, to my understanding at least, that at present we have no satisfactory "ultrafinitistic mathematics" (see the discussion here).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .