• Antikythera mechanistic simulation reimplemented in Rust (was: Re: on the nature of undecidability within computing: and refuting the church turing thesis)

    From Johann 'Myrkraverk' Oskarsson@johann@myrkraverk.invalid to comp.theory,comp.lang.misc on Mon Aug 31 19:23:43 2026
    From Newsgroup: comp.lang.misc

    On 31/08/2026 5:36 PM, Mikko wrote:
    On 31/08/2026 04:22, Johann 'Myrkraverk' Oskarsson wrote:
    On 28/08/2026 3:21 PM, Mikko wrote:
    On 27/08/2026 14:46, Johann 'Myrkraverk' Oskarsson wrote:
    On 27/08/2026 3:49 PM, Mikko wrote:
    On 27/08/2026 08:40, dart200 wrote:
    title of my next paper is tentative, but i'm kinda liking it. yes >>>>>> i'm quite serious about refuting the church turing thesis. i
    demonstrate how an idealized human agent can compute that which is >>>>>> not turing computable.

    Is there any way to prove that humans can compute anyhing not Turing >>>>> computable? Much can be computed with a Turing computable partial
    method.

    I previously gave this practical counter example, and you're welcome
    to prove me wrong.

    You pick up an Antikythera mechanism, and use it to compute something, >>>> anything.

    Unlikely, as the mechanism is not fylly known. Only one damaged example
    is known and there is no evicence that more was ever constructed.

    Dear Mikko,

    I'm afraid you're wading in error and confusion.-a Please stop breathing
    those sauna fumes in Finland, and do your original research.-a Here is
    mine.

    -a-a https://www.antikytheramechanism.co.uk/category/all-products

    So, 1) the mechanism is fully known, 2) and there have been reconstruct-
    ions since.-a Now, if you want to /define/ a reconstruction as an invalid
    method of arguing on Usenet, I'll just have to tip my top hat to you,
    and insist you invent a time machine instead.

    The reconstruction is reasonble but it is impossible to verify every
    detail. But the mechanism is understood sufficiently well that the
    the purpose and idea can be understood. That a part of the writing on
    the machine has been read also helps.

    The remaining uncertainties are small and irrelevant to our discussion.

    Now, after that feat, how do you do this with a Turing machine?
    -aYou can obviously simulate the Antikythera mechanism with a
    Turing> machine, but you cannot /compute/ as one.

    Simulation is enough. If one can simulate a machine one can compute
    what the simulated machine can compute.

    Let there be epsilon less than delta.-a Once you have constructed a simu-
    lation, I'll construct a level, and a fine bolt to adjust it.-a I believe
    I'll try for twenty thousand threads per inch.-a So the question is, can
    I nudge an arbitrary long lever, with a bolt -- hopefully starting at
    twenty thousand threads per inch -- finely enough, that each nudge fits
    within the smallest floating point precision you have -- I'm going to
    assume you were going to use double precision in C, REAL*8 in Fortran 77
    or f64 in Rust -- and you won't /see it/ because your simulation lacks
    real world precision?

    The topic was Turing computability. Particuar floating point limitations
    are irrelevant. Computability of real numbers means computability to any desired precision.


    Right, so you want to stick to that argument? Can you construct arbi-
    trary precision floating point library, and can I still create an arbi-
    trary long lever, with an arbitrary fine threaded bolt to push it, so
    that when you try to allocate more R.A.M. for the results, you panic
    because your tower of computation runs out of memory and forgets every-
    thing?

    I'm assuming you're going to write your simulator in Rust, being a duly anointed philosopher of a doctorate, I have no idea which, so I've added comp.lang.misc to this discussion, as comp.lang.rust is still being de-
    bated in news.groups.proposals.

    I will now await for you to finish said simulator, before I begin con- structing my very own mechanical shop of machinery, and /define/ myself
    as having won said argument until you show an actual working simulator
    I can then crash with a panic!


    Enjoy making the simulator in Rust!
    --
    Johann | email: invalid -> com | http://www.myrkraverk.com/blog/
    I'm not from the Internet, I just work there. | via Easynews.com https://bsky.app/profile/myrkraverk.bsky.social | for ( ;; ) _:;
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ross Finlayson@ross.a.finlayson@gmail.com to comp.theory,comp.lang.misc on Mon Aug 31 07:33:34 2026
    From Newsgroup: comp.lang.misc

    On 08/31/2026 04:23 AM, Johann 'Myrkraverk' Oskarsson wrote:
    On 31/08/2026 5:36 PM, Mikko wrote:
    On 31/08/2026 04:22, Johann 'Myrkraverk' Oskarsson wrote:
    On 28/08/2026 3:21 PM, Mikko wrote:
    On 27/08/2026 14:46, Johann 'Myrkraverk' Oskarsson wrote:
    On 27/08/2026 3:49 PM, Mikko wrote:
    On 27/08/2026 08:40, dart200 wrote:
    title of my next paper is tentative, but i'm kinda liking it. yes >>>>>>> i'm quite serious about refuting the church turing thesis. i
    demonstrate how an idealized human agent can compute that which
    is not turing computable.

    Is there any way to prove that humans can compute anyhing not Turing >>>>>> computable? Much can be computed with a Turing computable partial
    method.

    I previously gave this practical counter example, and you're welcome >>>>> to prove me wrong.

    You pick up an Antikythera mechanism, and use it to compute something, >>>>> anything.

    Unlikely, as the mechanism is not fylly known. Only one damaged example >>>> is known and there is no evicence that more was ever constructed.

    Dear Mikko,

    I'm afraid you're wading in error and confusion. Please stop breathing
    those sauna fumes in Finland, and do your original research. Here is
    mine.

    https://www.antikytheramechanism.co.uk/category/all-products

    So, 1) the mechanism is fully known, 2) and there have been reconstruct- >>> ions since. Now, if you want to /define/ a reconstruction as an invalid >>> method of arguing on Usenet, I'll just have to tip my top hat to you,
    and insist you invent a time machine instead.

    The reconstruction is reasonble but it is impossible to verify every
    detail. But the mechanism is understood sufficiently well that the
    the purpose and idea can be understood. That a part of the writing on
    the machine has been read also helps.

    The remaining uncertainties are small and irrelevant to our discussion.

    Now, after that feat, how do you do this with a Turing machine?
    You can obviously simulate the Antikythera mechanism with a
    Turing> machine, but you cannot /compute/ as one.

    Simulation is enough. If one can simulate a machine one can compute
    what the simulated machine can compute.

    Let there be epsilon less than delta. Once you have constructed a simu- >>> lation, I'll construct a level, and a fine bolt to adjust it. I believe >>> I'll try for twenty thousand threads per inch. So the question is, can
    I nudge an arbitrary long lever, with a bolt -- hopefully starting at
    twenty thousand threads per inch -- finely enough, that each nudge fits
    within the smallest floating point precision you have -- I'm going to
    assume you were going to use double precision in C, REAL*8 in Fortran 77 >>> or f64 in Rust -- and you won't /see it/ because your simulation lacks
    real world precision?

    The topic was Turing computability. Particuar floating point limitations
    are irrelevant. Computability of real numbers means computability to any
    desired precision.


    Right, so you want to stick to that argument? Can you construct arbi-
    trary precision floating point library, and can I still create an arbi-
    trary long lever, with an arbitrary fine threaded bolt to push it, so
    that when you try to allocate more R.A.M. for the results, you panic
    because your tower of computation runs out of memory and forgets every- thing?

    I'm assuming you're going to write your simulator in Rust, being a duly anointed philosopher of a doctorate, I have no idea which, so I've added comp.lang.misc to this discussion, as comp.lang.rust is still being de-
    bated in news.groups.proposals.

    I will now await for you to finish said simulator, before I begin con- structing my very own mechanical shop of machinery, and /define/ myself
    as having won said argument until you show an actual working simulator
    I can then crash with a panic!


    Enjoy making the simulator in Rust!

    "Interval arithmetic" is a usual sort of principled account
    with regards to "extended-precision arithmetic" that though
    mostly the "computer algebra systems" keep things formulaic
    throughout and only approximate numbers at the end.

    It's like when computing rotations in computer graphics,
    and there's a great account that matrix-manipulation the
    matrix-manipulation starts to have that 90-degree rotations
    start looking slightly off, i.e., it's relevant to getting
    things straight left-to-right and up-to-down, or according
    to the various coordinate and sign conventions or representations
    on a grid of picture-elements, that for simple accounts like
    vertical and horizontal lines on paper that making floating-point
    is graphically unsettling.


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Johann 'Myrkraverk' Oskarsson@johann@myrkraverk.invalid to comp.theory,comp.lang.misc on Tue Sep 1 03:58:50 2026
    From Newsgroup: comp.lang.misc

    On 31/08/2026 10:33 PM, Ross Finlayson wrote:
    On 08/31/2026 04:23 AM, Johann 'Myrkraverk' Oskarsson wrote:
    On 31/08/2026 5:36 PM, Mikko wrote:
    On 31/08/2026 04:22, Johann 'Myrkraverk' Oskarsson wrote:
    On 28/08/2026 3:21 PM, Mikko wrote:
    On 27/08/2026 14:46, Johann 'Myrkraverk' Oskarsson wrote:
    On 27/08/2026 3:49 PM, Mikko wrote:
    On 27/08/2026 08:40, dart200 wrote:
    title of my next paper is tentative, but i'm kinda liking it. yes >>>>>>>> i'm quite serious about refuting the church turing thesis. i
    demonstrate how an idealized human agent can compute that which >>>>>>>> is not turing computable.

    Is there any way to prove that humans can compute anyhing not Turing >>>>>>> computable? Much can be computed with a Turing computable partial >>>>>>> method.

    I previously gave this practical counter example, and you're welcome >>>>>> to prove me wrong.

    You pick up an Antikythera mechanism, and use it to compute
    something,
    anything.

    Unlikely, as the mechanism is not fylly known. Only one damaged
    example
    is known and there is no evicence that more was ever constructed.

    Dear Mikko,

    I'm afraid you're wading in error and confusion.-a Please stop breathing >>>> those sauna fumes in Finland, and do your original research.-a Here is >>>> mine.

    -a-a https://www.antikytheramechanism.co.uk/category/all-products

    So, 1) the mechanism is fully known, 2) and there have been
    reconstruct-
    ions since.-a Now, if you want to /define/ a reconstruction as an
    invalid
    method of arguing on Usenet, I'll just have to tip my top hat to you,
    and insist you invent a time machine instead.

    The reconstruction is reasonble but it is impossible to verify every
    detail. But the mechanism is understood sufficiently well that the
    the purpose and idea can be understood. That a part of the writing on
    the machine has been read also helps.

    The remaining uncertainties are small and irrelevant to our discussion.

    Now, after that feat, how do you do this with a Turing machine?
    You can obviously simulate the Antikythera mechanism with a
    Turing> machine, but you cannot /compute/ as one.

    Simulation is enough. If one can simulate a machine one can compute
    what the simulated machine can compute.

    Let there be epsilon less than delta.-a Once you have constructed a
    simu-
    lation, I'll construct a level, and a fine bolt to adjust it.-a I
    believe
    I'll try for twenty thousand threads per inch.-a So the question is, can >>>> I nudge an arbitrary long lever, with a bolt -- hopefully starting at
    twenty thousand threads per inch -- finely enough, that each nudge fits >>>> within the smallest floating point precision you have -- I'm going to
    assume you were going to use double precision in C, REAL*8 in
    Fortran 77
    or f64 in Rust -- and you won't /see it/ because your simulation lacks >>>> real world precision?

    The topic was Turing computability. Particuar floating point limitations >>> are irrelevant. Computability of real numbers means computability to any >>> desired precision.


    Right, so you want to stick to that argument?-a Can you construct arbi-
    trary precision floating point library, and can I still create an arbi-
    trary long lever, with an arbitrary fine threaded bolt to push it, so
    that when you try to allocate more R.A.M. for the results, you panic
    because your tower of computation runs out of memory and forgets every-
    thing?

    I'm assuming you're going to write your simulator in Rust, being a duly
    anointed philosopher of a doctorate, I have no idea which, so I've added
    comp.lang.misc to this discussion, as comp.lang.rust is still being de-
    bated in news.groups.proposals.

    I will now await for you to finish said simulator, before I begin con-
    structing my very own mechanical shop of machinery, and /define/ myself
    as having won said argument until you show an actual working simulator
    I can then crash with a panic!


    Enjoy making the simulator in Rust!

    "Interval arithmetic" is a usual sort of principled account
    with regards to "extended-precision arithmetic" that though
    mostly the "computer algebra systems" keep things formulaic
    throughout and only approximate numbers at the end.

    Do you mean to say that an /Antikythera mechanical simulator/
    can be done purely in algebraic terms, and does not need float-
    ing point at all?

    Wouldn't that simulator just run out of R.A.M. faster than arbi-
    trary precision floating point library?


    It's like when computing rotations in computer graphics,
    and there's a great account that matrix-manipulation the
    matrix-manipulation starts to have that 90-degree rotations
    start looking slightly off, i.e., it's relevant to getting
    things straight left-to-right and up-to-down, or according
    to the various coordinate and sign conventions or representations
    on a grid of picture-elements, that for simple accounts like
    vertical and horizontal lines on paper that making floating-point
    is graphically unsettling.



    I thought this was already solved in the 90s -- or earlier? -- by
    using quaternions instead of matrices for the rotations, and that
    it would remove the weird wobbling that comes from repeated calcu-
    lations of the rotation matrix?


    Please enlighten us on this subject!
    --
    Johann | email: invalid -> com | http://www.myrkraverk.com/blog/
    I'm not from the Internet, I just work there. | via Easynews.com https://bsky.app/profile/myrkraverk.bsky.social | for ( ;; ) _:;
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ross Finlayson@ross.a.finlayson@gmail.com to comp.theory,comp.lang.misc on Mon Aug 31 20:22:59 2026
    From Newsgroup: comp.lang.misc

    On 08/31/2026 12:58 PM, Johann 'Myrkraverk' Oskarsson wrote:
    On 31/08/2026 10:33 PM, Ross Finlayson wrote:
    On 08/31/2026 04:23 AM, Johann 'Myrkraverk' Oskarsson wrote:
    On 31/08/2026 5:36 PM, Mikko wrote:
    On 31/08/2026 04:22, Johann 'Myrkraverk' Oskarsson wrote:
    On 28/08/2026 3:21 PM, Mikko wrote:
    On 27/08/2026 14:46, Johann 'Myrkraverk' Oskarsson wrote:
    On 27/08/2026 3:49 PM, Mikko wrote:
    On 27/08/2026 08:40, dart200 wrote:
    title of my next paper is tentative, but i'm kinda liking it. yes >>>>>>>>> i'm quite serious about refuting the church turing thesis. i >>>>>>>>> demonstrate how an idealized human agent can compute that which >>>>>>>>> is not turing computable.

    Is there any way to prove that humans can compute anyhing not
    Turing
    computable? Much can be computed with a Turing computable partial >>>>>>>> method.

    I previously gave this practical counter example, and you're welcome >>>>>>> to prove me wrong.

    You pick up an Antikythera mechanism, and use it to compute
    something,
    anything.

    Unlikely, as the mechanism is not fylly known. Only one damaged
    example
    is known and there is no evicence that more was ever constructed.

    Dear Mikko,

    I'm afraid you're wading in error and confusion. Please stop
    breathing
    those sauna fumes in Finland, and do your original research. Here is >>>>> mine.

    https://www.antikytheramechanism.co.uk/category/all-products

    So, 1) the mechanism is fully known, 2) and there have been
    reconstruct-
    ions since. Now, if you want to /define/ a reconstruction as an
    invalid
    method of arguing on Usenet, I'll just have to tip my top hat to you, >>>>> and insist you invent a time machine instead.

    The reconstruction is reasonble but it is impossible to verify every
    detail. But the mechanism is understood sufficiently well that the
    the purpose and idea can be understood. That a part of the writing on
    the machine has been read also helps.

    The remaining uncertainties are small and irrelevant to our discussion. >>>>
    Now, after that feat, how do you do this with a Turing machine? >>>>>> > You can obviously simulate the Antikythera mechanism with a
    Turing> machine, but you cannot /compute/ as one.

    Simulation is enough. If one can simulate a machine one can compute >>>>>> what the simulated machine can compute.

    Let there be epsilon less than delta. Once you have constructed a
    simu-
    lation, I'll construct a level, and a fine bolt to adjust it. I
    believe
    I'll try for twenty thousand threads per inch. So the question is,
    can
    I nudge an arbitrary long lever, with a bolt -- hopefully starting at >>>>> twenty thousand threads per inch -- finely enough, that each nudge
    fits
    within the smallest floating point precision you have -- I'm going to >>>>> assume you were going to use double precision in C, REAL*8 in
    Fortran 77
    or f64 in Rust -- and you won't /see it/ because your simulation lacks >>>>> real world precision?

    The topic was Turing computability. Particuar floating point
    limitations
    are irrelevant. Computability of real numbers means computability to
    any
    desired precision.


    Right, so you want to stick to that argument? Can you construct arbi-
    trary precision floating point library, and can I still create an arbi-
    trary long lever, with an arbitrary fine threaded bolt to push it, so
    that when you try to allocate more R.A.M. for the results, you panic
    because your tower of computation runs out of memory and forgets every-
    thing?

    I'm assuming you're going to write your simulator in Rust, being a duly
    anointed philosopher of a doctorate, I have no idea which, so I've added >>> comp.lang.misc to this discussion, as comp.lang.rust is still being de-
    bated in news.groups.proposals.

    I will now await for you to finish said simulator, before I begin con-
    structing my very own mechanical shop of machinery, and /define/ myself
    as having won said argument until you show an actual working simulator
    I can then crash with a panic!


    Enjoy making the simulator in Rust!

    "Interval arithmetic" is a usual sort of principled account
    with regards to "extended-precision arithmetic" that though
    mostly the "computer algebra systems" keep things formulaic
    throughout and only approximate numbers at the end.

    Do you mean to say that an /Antikythera mechanical simulator/
    can be done purely in algebraic terms, and does not need float-
    ing point at all?

    Wouldn't that simulator just run out of R.A.M. faster than arbi-
    trary precision floating point library?


    It's like when computing rotations in computer graphics,
    and there's a great account that matrix-manipulation the
    matrix-manipulation starts to have that 90-degree rotations
    start looking slightly off, i.e., it's relevant to getting
    things straight left-to-right and up-to-down, or according
    to the various coordinate and sign conventions or representations
    on a grid of picture-elements, that for simple accounts like
    vertical and horizontal lines on paper that making floating-point
    is graphically unsettling.



    I thought this was already solved in the 90s -- or earlier? -- by
    using quaternions instead of matrices for the rotations, and that
    it would remove the weird wobbling that comes from repeated calcu-
    lations of the rotation matrix?


    Please enlighten us on this subject!


    I don't know that off-hand, it's an outstanding problem and well-known limitation of finite-element-analysis, and about the imprecision of
    floating points, and the incompleteness of rational arithmetic,
    about rational-algebras and radical-magmas.

    The gimbal-lock is an error mode in systems of gyroscopes, where
    for example "that did gyre and gimble in the wabe", whereas the
    quaternions don't suffer that themselves, and help avoid "lerps"
    and the like, then in the setting of where rotation-matrices are
    used to indicate usually in a wider setting the translation, rotation,
    and skew of the affine, or transformation-matrices in computer graphics,
    then usually providing rational rotations of multiples
    of pi/2 radians just involves constant cases since as arithmetic
    would accumulate its imprecision it would be off.

    Then, for seconds, days, and years, and clock-arithmetic, then
    a usual idea is that there are separate sorts time-keeping methods
    for each, then its similar for systems as would be like an analog
    computer that involves things like the _yenri_, which is a form
    of calculus from Japan that's about circles-within-circles as
    for epicycles and the like instead of the quadratic and the usual
    account of trapezoid rule and integration the the integral calculus,
    then usually the idea is to compute the time-of-day then draw the
    watch-hands, which in reality, actually tick on through a continuous
    sweep of motion.



    --- Synchronet 3.22a-Linux NewsLink 1.2