1,000
3.00@ArmandodiMatteo see my comments below for more info and a possibly less ambiguous market.
But for the current record holder, we have formulas that can be written in a human readable way, using only the basic arithmetic operations, parentheses, and exponentiation, that provide a value, and the lower and upper bounds are identical.
That's probably as good as you can get for a number this size.
For the current record holder (the Kroptiz 2022 t15 machine), the exact number of steps taken and marks made is known, in the sense that there exists a precise formula. The best lower bound and best upper bound are identical; it's merely "> 10↑↑15" because that's a compact lower bound that's easily written in text.
https://www.sligocki.com/2022/06/21/bb-6-2-t15.html
So, if /EvanDaniel/is-the-bb6-machine-the-kropitz-2022 is true and proven so in time, this probably should resolve true as well, depending on how you feel about the existence of large and precisely described numbers.
Also relevant:
https://bbchallenge.org/~pascal.michel/bbc
Can we produce a collection of statements that might mean we "know" the number?
We know explicitly which machine is the BB. (This seems uncontroversially required for a 'YES' resolution, right? I suppose we could find ourselves in an awkward state of having proven that the BB is one of a small set of machines, all proven to halt, but not having a proof of which one halts first. That would provide a finite but impractical procedure to generate the number.)
We have both a lower and upper bound, expressed in some fashion more comprehensible than "the number output by this TM". (For example, a proof that the number is between 10↑↑15 and 10↑↑16.) (Is there some clean definition for "comprehensible" here? Something primitive recursive?)
We know some specific digits of the number (in some base, not necessarily base 10).
We can construct a program that outputs the digits in [reverse?] order, at some reasonably-bounded interval between successive digits generated. (Or some other formalization of "the program does something more interesting than run the known machine until it halts and then outputs the number".) This might be the closest to a "constructivist" definition of "know" that applies to a number that large.
We can do some sort of math / arithmetic with the number. For example, we can prove some prime factors of the number (or even that some primes are not factors).
We can compactly describe some large portion of the final tape. (This probably says more about understanding the behavior of the machine than the number itself.)
Other suggestions? Are there other criteria that feel like they get close to a constructivist definition of "know"?
Under any usual definition of "knowing the exact value of a natural number", you can already resolve to NO.
Even with the ultra-weak definition of knowing a number by having a Turing machine that will eventually write the digits, NO is a sure bet.
You can have an AGI, superintelligent AI, the singularity, rapture of the nerds, moving the Earth as far as you want from the Sun but you will never know the exact value of Σ(6)
@Zardoru having a turing machine that will eventually write the digits is possible. we have a turing machine which will eventually write the digits of Graham's number, for example.
@SeraphinaNix I didn't mention this yesterday, but there are additional complications. Suppose we are eventually able to decide every machine except for the candidate and one other that provably runs for longer. Suppose the question of whether this other machine halts has not been explicitly proved, but has been proved to be decidable (that is, we know an explicit Turing machine which returns whether or not it halts). Then we can construct a Turing machine which we know outputs the Busy Beaver number, but we cannot in good conscience make a statement of the form "The Busy Beaver machine is X".
@BoltonBailey that's fair. there's actually a simpler construction with the same problem, which is just dovetail all the 6 state turing machines, and output the largest one that has halted so far. Both constructions never halt, and both constructions will eventually output the correct answer.
@SeraphinaNix I think there is a pretty good "know it when I see it" resolution criteria here, though if we want to formalize it, I think asking for a proof that a specific and explicit turing machine is the BB champion for size 6 is both unambiguous and what the spirit of the question is asking.
@SeraphinaNix Ok your davetail machine eventually output the answer, however it never halt and when it output an anwser we can't know if it will be the last and correct one or if there will be another one later.
@Zardoru That is true of course. I thought that was true of Bolton's machine as well, but it's actually not. I, however, think Bolton's machine (now that I've reread the definition) is rather unlikely to be our state of knowledge. However, if it was, I don't think that should be sufficient by itself to resolve this market YES.
@Zardoru We have a Turing machine that would eventually write the digits. We just can’t know if it will be implemented and it doesn’t halt.
Going by the "we know which machine it is" definition, I think there's still a strong NO case. The current contender for 5 was discovered in 1990, and has not been proven to be the actual Busy Beaver. The current contender for 6 was discovered last year, and I would expect to take much longer to prove, if it's even possible in our most-used models of mathematics.
@IsaacKing Given the information on the Wikipedia page, this value has more digits than there are Planck volumes in the observable universe. I would philosophically question any definitions of "exact" and "known" which would cause this market could resolve YES.
@IsaacKing What is your definition of "we know the exact value of this number"? Is it "we have in hand a description of a Turing machine which we know will output this number"?
@BoltonBailey Hmm. If we know the specific machine and could in principle run it, I think that should count.
@IsaacKing I'll acknowledge that this is a valid definition, but I'll still express skepticism about it being a good definition for "know". You say we know the exact value of Graham's number, would you say we know the exact value if its first digit in base 10?
@BoltonBailey Would you say we know the exact value of pi, even though we don't know its quadrillionth digit? How would you define "knowing the exact value of a number" in a way that doesn't privilege an arbitrary notation method?
@IsaacKing I would probably take a finitist approach and say that we know the value of a number if we can write it down in binary or decimal. Indeed, that might be privileging binary over unary, but it feels appropriate since I think this is how humans normally deal with numbers.
I would probably say we don't know the exact value of pi. Do you think we do? The first google result I get for "do we know the exact value of pi" is "The pi is an irrational number and does not have an exact value." Note that it doesn't just say we don't know it, it says it doesn't even exist which I think lines up with my approach of can we write it down.
I would say that we do know the exact value of a rational number if we know the exact value of the numerator and denominator. I think I just go by "whatever the most standard way of writing a value of that type down is".
The pi is an irrational number and does not have an exact value.
Congratulations, you've managed to find someone with beliefs I find truly incomprehensible. People who think the Earth is flat? Yeah ok, I see how they got there. But this‽ This is true alien thought.
Of course it has an exact value! It's a single point on the real number line just like any other real number!
I would say that we do know the exact value of a rational number if we know the exact value of the numerator and denominator.
Cool, Graham's Number / 1. Glad we settled that. :)
Yes, I'll admit that the "real numbers aren't real" is an uncommon position, but it is one that I hold. In physics do we ever talk about real numbers, or do we just talk about rational numbers with rational error bars? Do we ever really use pi, or is pi just a mathematical shorthand for a sequence of approximations that can be algebraically manipulated like a number?
do we ever really talk about the continuum, or do we only ever talk about finite sequences of symbols that talk about the continuum?
@BoltonBailey That's interesting : we know a lot of last digits of Graham's number (https://oeis.org/A133613) but not the very first.
@BoltonBailey π is in fact a number, just like s(s(0)). It’s half of the ratio of the circumference of a circle to its radius, and you don’t see anyone claiming that circumferences or radii aren’t numbers.