[ExI] Gemini on Mathematical Realism

Wed Oct 7 17:50:06 UTC 2026

On Tue, Oct 6, 2026 at 6:23 PM Stuart LaForge via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> On 2026-10-06 14:31, Jason Resch via extropy-chat wrote:
>
> [snip]

> > So then where does constructivism enter in this picture concerning the
> > objective truths of Turing machines?
>
> I fed your prompt into Gemini:
>
> Jason Resch: The halting (or not halting) of various Turing machines is
> by all accounts
> objective, a given machine given all the time it needs will either halt,
> or
> it won't. This creates a set of objective truth that no finite axiomatic
> system can capture. Different axiomatic systems will be a able to prove
> more things about more machines than others and hence will be
> objectively more powerful systems. So then where does constructivism
> enter in this picture concerning the objective truths of Turing
> machines?
>
> Gemini: Constructivism enters this picture at three distinct points:
> **the challenge to the Law of Excluded Middle**, **the asymmetry of
> proof witnesses**, and **the finite construction of non-halting
> invariants**.
>
> While the step-by-step state transition of a Turing machine is an
> objective, deterministic process, our *epistemological access* to its
> infinite behavior depends entirely on constructive tools.
>

I agree on this. For example, we can't calculate the Googolplexth decimal
digit of Pi in this universe. But I would still say there is nevertheless
an objective fact concerning its particular value. If all of Pi's infinite
digits did not exist, and have the exact values they happen to have, then
e^(2 * Pi * i) would not equal 1. The fact that someone born in a more
impoverished than ours might lack the resources to compute the trillionth
digit of Pi, but their limitation says nothing about Pi's digits not having
particular objective values. Does Gemini agree with this? If someone in a
universe with greater memory and computational resources than ours does
compute the Googolplexth decimal digit of Pi, does that make it exist for
everyone across reality? Or are mathematical truths observer- and
universe-relative?

>
>
> ### 1. The Constructivist Critique: Challenging $Halt(M) \lor \neg
> Halt(M)$
>
> The premise that *"a machine given all the time it needs will either
> halt or it won't"* relies on classical logic's **Law of Excluded Middle
> (LEM)** applied across an infinite domain:
>
> $$\forall M \left( \text{Halt}(M) \lor \neg \text{Halt}(M) \right)$$
>
> Constructivists and Intuitionists (from L.E.J. Brouwer to Errett Bishop)
> reject the blanket application of LEM to infinite or unbounded search
> spaces.
>
> * **The Constructive Rule:** A constructivist does not accept $P \lor
> \neg P$ as an automatic truth unless you possess an effective, finite
> procedure to determine *which* of the two disjuncts holds.
> * **The Implication:** Because Alan Turing proved that no general
> decision algorithm exists (the Halting Problem), a strict constructivist
> argues that "non-halting" is not an abstract metaphysical fact floating
> in space—it only becomes a mathematical truth when an explicit proof or
> invariant can be **constructed**.
>

Is this a view that Gemini is defending? If you require realization then in
what sense is your hybrid view realist? My understanding of the hybrid
position is that the infinite consequences exist and are objective, but
there must first be some initial act of specifying the particular formal
system. Is this an accurate description of the hybrid view?

>
> ---
>
> ### 2. The Asymmetry of Witnesses: Concrete Trace vs. Constructed
> Invariant
>
> The physical nature of computation highlights a deep structural
> asymmetry between halting and non-halting:
>
> * **Halting is Naturally Constructive:** If a machine halts at step $N$,
> the truth of its halting is certified by a concrete, finite witness—the
> state trace of $N$ steps. You don't need faith or high-level axioms; you
> simply run the machine and observe step $N$.
> * **Non-Halting Requires Structural Construction:** You cannot "observe"
> non-halting by simply waiting, because at step $N + 1,000,000$, the
> machine might still halt. To establish that a machine *never* halts, you
> cannot rely on empirical execution. You must **construct a finite
> symbolic proof**—such as an inductive invariant showing that a specific
> memory state can never recur, or a Lyapunov function on the tape head.
>
> Non-halting facts only enter our sphere of knowledge when we construct a
> formal argument that traps the machine's infinite trajectory inside a
> finite logical box.
>

Indeed, but does this provide a means for comparing ever more powerful
axiomatic systems? Is system A better than system B because A can prove
that certain machines halt, which system B cannot prove?

>
> ---
>
> ### 3. How This Fits the "Formalism in Construct, Realism in
> Consequence" Stance
>
> This brings us back to the synthesis between constructed tools and
> objective landscapes:
>
> ```
> [ Constructed Rule Engine ]            [ Objective Execution ]
>     [ Constructed Proof Witness ]
> Turing Machine Definition   --->   Step-by-step State Transitions   --->
>    Inductive Invariant / Proof
>    (Syntax & Alphabet)                 (Deterministic Reality)
>       (Constructive Access)
>
> ```
>
> 1. **The Machine is Constructed:** The definition of a tape, alphabet,
> and state transition function $T(s, a) \to (s', a', D)$ is an
> engineered, formal system.
> 2. **The Execution is Real:** Once those rules are set, the sequence of
> machine configurations $C_0 \to C_1 \to C_2 \dots$ is completely
> deterministic and mind-independent.
> 3. **"More Powerful" Systems are Constructed Upgrades:** When we say
> System $B$ (e.g., $ZFC + \text{Large Cardinals}$) is "more powerful"
> than System $A$ (e.g., Peano Arithmetic) because it resolves the halting
> status of the Busy Beaver function $BB(n)$ for larger $n$, where did
> that power come from?
> * It didn't come from discovering a Platonic object. It came from
> **constructing a richer axiomatic framework** with stronger reflection
> principles ($Con(A)$), allowing us to build shorter, stronger proof
> witnesses for non-halting machines.
>
>
>
> The objective landscape of computation is deterministic, but it is
> infinite. Constructivism is the requirement that any path we map through
> that infinite landscape must be paved with finite, verifiable symbolic
> steps.
>

Once a given formal system is defined, what status do you ascribe to the
propositions that are true and provable in that system, but to which:

1. No one has yet bothered to prove
2. It was proven, but then all records of that proof are erased and lost
3. Someone has proved it, but it was in a location causally isolated from
the location under consideration (e.g. far beyond the cosmological horizon)
4. The given universe lacks the resources to ever prove it
5. Someone has proved it, but only in a different universe within a larger
multiverse of causally isolated universes.

> -----------
>
> Wow. Feel free to push back on Gemini again. I am going to sit this one
> out. ;)
>
>
Challenge accepted. :-)

Jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20261007/d82cde4f/attachment-0001.htm>