• Re: What value should HHH(DD) correctly return?

    From Richard Heathfield@21:1/5 to olcott on Wed Aug 6 19:57:28 2025
    XPost: comp.theory

    On 06/08/2025 19:34, olcott wrote:
    Simulating Termination Analyzer HHH correctly simulates its input

    No, it doesn't.

    You posted your question in comp.lang.c, so I will post a C
    answer in comp.lang.c and honour your followup request by copying
    it to comp.theory, but it is a C question belonging in a C group,
    so I have set followups to comp.lang.c.

    In C, HHH as defined cannot correctly simulate its input. You
    declare HHH as taking a pointer to function returning void, but
    you attempt to pass it a pointer to a function returning int.

    If we gloss over that oversight, we run up against the problem
    that while C allows you to call DD from HHH by using
    pointer-deferencing, it does not offer you the access you seek to
    the machine code instructions that make up the function's object
    code. A given implementation might offer such access as an
    extension, but C does not impose upon any implementation any
    obligation to provide such access as you seek, and your attempt
    to take advantage of it takes you firmly into the realm of
    undefined behaviour.

    In short, your question is based on a misunderstanding of the C
    language, and the only meaningful answer is that when you break
    the rules of C, all bets are off.

    <snip>

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Wed Aug 6 22:23:57 2025
    On 06/08/2025 21:40, olcott wrote:
    On 8/6/2025 1:57 PM, Richard Heathfield wrote:
    On 06/08/2025 19:34, olcott wrote:
    Simulating Termination Analyzer HHH correctly simulates its input

    No, it doesn't.

    <snip>

    If we gloss over that oversight, we run up against the problem
    that while C allows you to call DD from HHH by using
    pointer-deferencing, it does not offer you the access you seek
    to the machine code instructions that make up the function's
    object code. A given implementation might offer such access as
    an extension, but C does not impose upon any implementation any
    obligation to provide such access as you seek, and your attempt
    to take advantage of it takes you firmly into the realm of
    undefined behaviour.


    int DD()
    {
      int Halt_Status = HHH(DD);
      if (Halt_Status)
        HERE: goto HERE;
      return Halt_Status;
    }

    int main()
    {
      unsigned char *P;
      __asm mov eax, DD

    You're no longer writing in the language defined by ISO. Whether
    it does what you want depends entirely on whether it does what
    you want. The rules of C do not impose any requirements on this code.

      __asm mov P, eax
      for (int N = 0; N < 35; N++)
        printf("[%02d]  %02x \n", N, (unsigned char) *(P+N));
    }

    Not standard C but it works on ever compiler
    that enables embedded assembly language.

    By which I presume you mean the one or two you've tried it on.

    Your question (subject line) is: "What value should HHH(DD)
    correctly return?"

    And I have answered that question: "In C, HHH as defined cannot
    correctly simulate its input."

    If you are now changing the question to "does it work on the
    platforms it works on?" it's no longer a question about the C
    language, so I suggest you ask it in a newsgroup devoted to the
    platform you care about.

    In short, your question is based on a misunderstanding of the C
    language, and the only meaningful answer is that when you break
    the rules of C, all bets are off.

    <snip>


    So that is just a limitation of C implementations

    On the contrary! ISO C grants C implementations licence to
    implement any extension they wish. But it doesn't force all C
    implementations to implement all the extensions offered by other
    C implementations. The core C language and library are portable,
    but if you step outside the core (as you have done) the language
    no longer defines what your program will do.

    that does not actually negate the gist of my reasoning.

    What negates the gist of your reasoning is that the behaviour of
    your code is not defined by the rules of C.

    What your code does on any given run can be determined by
    observation. If it does the same thing on the next run, count
    yourself lucky.


    My x86utm operating system

    may be topical in an X86 newsgroup. This is not that group.

    Then on the basis of the correct emulation of HHH(DD)

    Stop right there.

    In comp.lang.c we discuss C, not pipedreams. HHH cannot correctly
    emulate anything passed to it as a function pointer. It can call
    the function, but that's all.

    <pipedream snipped>

    It is a proven fact that HHH(DD) does correctly emulate

    That's not what the rules of C tell us.

    You are confusing implementation extensions with the rules of C.
    They are very different things.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Wed Aug 6 23:21:11 2025
    On 06/08/2025 23:11, olcott wrote:
    On 8/6/2025 4:23 PM, Richard Heathfield wrote:
    On 06/08/2025 21:40, olcott wrote:

    <snip>

    It is a proven fact that HHH(DD) does correctly emulate

    That's not what the rules of C tell us.

    You are confusing implementation extensions with the rules of
    C. They are very different things.


    This does work in C/C++

    No. You don't get to decide what those terms mean.

    You have the correct answer to your question.

    If you don't like the answer, that really is your problem. Asking
    the question again and posting your code again won't change the
    answer.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Thu Aug 7 00:13:02 2025
    On 06/08/2025 23:50, olcott wrote:
    On 8/6/2025 5:21 PM, Richard Heathfield wrote:

    <snip>


    You have the correct answer to your question.

    If you don't like the answer, that really is your problem.
    Asking the question again and posting your code again won't
    change the answer.


    If we assume away

    But we don't.

    that limitation of C and simply understand
    that there is some complex infrastructure that does handle
    that then my 22 years of work can be easily validated.

    If you assume you're right, in your eyes you're right.

    The rest of us are not as gullible.

    At least now with your help I do know that I must at least
    reference this underlying infrastructure.

    It won't help with your goal, because your whole simulation thing
    is just a blind alley.

    How valuable to computer science would it be to understand
    that the conventional halting problem proofs do not actually
    derive their undecidability conclusion?

    The conventional halting problem proof took a genius to figure
    out, but it's really quite trivial. I don't know of any
    intelligent programmer who struggles to understand why it's true.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kaz Kylheku@21:1/5 to Richard Heathfield on Thu Aug 7 00:44:47 2025
    On 2025-08-06, Richard Heathfield <[email protected]> wrote:
    On 06/08/2025 23:11, olcott wrote:
    On 8/6/2025 4:23 PM, Richard Heathfield wrote:
    On 06/08/2025 21:40, olcott wrote:

    <snip>

    It is a proven fact that HHH(DD) does correctly emulate

    That's not what the rules of C tell us.

    You are confusing implementation extensions with the rules of
    C. They are very different things.


    This does work in C/C++

    No. You don't get to decide what those terms mean.

    The main problem with Peter Olcott's reasoning is that he equivocates among multiple test cases as being one single program operating on one input.

    There is a kind of Cantor diagonal argument at play in Halting in that
    we have an infinite table with rows that are halting functions, and
    columns that are test cases. Entries in the table indicate the halting
    or not. Along the diagonal is where the trick test cases lie: procedures
    that behave opposite to the halting verdict.

    The undecidability of halting says that in every row of the table there
    is a wrongly decided or non-halting column.

    Peter doesn't understand this; he thinks that a halting decider that
    has been deceived can see through the ruse, mutate itself and correctly decide---without that test case being identified as different row/column
    column of the table!

    When Peter uses pronouns like "itself", "it" and whatnot, or even
    explicit names like HH or whatever, those terms all equivocate among
    multiple different entities in a way that suits the narrative at hand.

    He is not able to grasp the concept that a symbol must name exactly one, immutable mathematical object, and that association must not change over
    the course of a valid argument. If equivocation is perpetrated over
    even a single symbol, the whole train of reasoning likely falls victim
    to fallacy.

    Peter is deeply committed to these errors of reasoning, and
    that is why he doesn't get anywhere, and his debaters likewise.

    Least effective is the strategy of debating the points that arise out of
    the fallacious reasoning, and are several steps removed. When you do
    that, you are accepting the bad premises, and the debate then takes
    place on Peter's terms (or even if not in reality, in his mind it does).

    Peter created (and I believe this) a simulation system based
    on some off-the-shelf emulation code for the x86 architecture.
    In this sytem he can have functions tracing the execution of other
    functions at mulitiple nestings. He has used this system to "prove"
    things to himself. I've never had access to his system to run it, but
    from looking in detail at his execution traces, I found issues
    such as gaps: pauses in the log where the instrucction pointer suddenly
    jumps somewhere. His approach encodes the fallacious reasoning;
    for instance, he identifies a procedure to be decided for halting
    only by its address in the x86 address space, even though halting is
    a property of program/input pairs, not of programs. The code is
    befuddled by impurity such as global variables.

    Peter went to a considerable software development effort in order to
    translate his fallacious chains of reasoning into code, the idea being
    that if something is represented in nicely executing, clever code,
    its results must be speaking the truth, and it will be held in high
    esteem by computer scientists as quality work who will accept the
    conclusions. Or something like that.

    --
    TXR Programming Language: http://nongnu.org/txr
    Cygnal: Cygwin Native Application Library: http://kylheku.com/cygnal
    Mastodon: @[email protected]

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Thu Aug 7 04:35:28 2025
    On 07/08/2025 00:46, olcott wrote:
    On 8/6/2025 6:13 PM, Richard Heathfield wrote:
    On 06/08/2025 23:50, olcott wrote:

    <snip>

    At least now with your help I do know that I must at least
    reference this underlying infrastructure.

    It won't help with your goal, because your whole simulation
    thing is just a blind alley.


    Unless you spend five minutes carefully studying
    this single page of text.
    https://philpapers.org/archive/OLCHPS.pdf

    It only took a second or two to see that your C is compromised
    the same way there as in your OP.

    The C answer to your C question stands - your C code is broken
    and rests on a false premise. C doesn't work the way you want it to.

    The conventional halting problem proof took a genius to figure
    out, but it's really quite trivial. I don't know of any
    intelligent programmer who struggles to understand why it's true.


    *This is very difficult to understand*

    No, it really isn't. You're just wrong and your C sucks.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Thu Aug 7 06:01:31 2025
    On 07/08/2025 04:55, olcott wrote:
    Given that HHH can recursively simulate an instance of itself

    It can't. Not in comp.lang.c it can't.

    simulating an instance of DD: What value should HHH(DD) correctly
    return?

    It can't correctly return anything. The language doesn't follow
    your rules. If you want correctness, first you must fix the C
    code to follow C rules.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Thu Aug 7 06:08:24 2025
    On 07/08/2025 05:05, olcott wrote:
    On 8/6/2025 10:35 PM, Richard Heathfield wrote:

    <snip>

    The C answer to your C question stands - your C code is broken
    and rests on a false premise. C doesn't work the way you want
    it to.


    The infrastructure that enables one C function to simulate
    another C function in debugstep mode is provided by the
    x86utm operating system.

    Then maybe that's where your answer lies. Ask in an x86utm
    newsgroup. This is not that group.

    If there is no such group: lobby the Big 8 Committee, demonstrate
    to them the overwhelming demand for such a group, and have them
    create it for you. Then, when the x86utm experts flock to that
    group, you can ask them your question.

    Given that HHH can recursively
    simulate an instance of itself

    Not a given, not in C. You don't get to define away language
    boundaries.

    What value should HHH(DD) correctly return?

    Make it correct C first. Then maybe the question can become
    meaningful.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From James Kuyper@21:1/5 to olcott on Thu Aug 7 15:14:26 2025
    On 06/08/2025 23:50, olcott wrote:
    ...
    How valuable to computer science would it be to understand
    that the conventional halting problem proofs do not actually
    derive their undecidability conclusion?

    Understanding a falsehood to be true would have a very large negative
    value. The validity of the proof of the halting problem isn't hard for
    most mathematically minded people to understand - the hard part was
    coming up with it.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Thu Aug 7 21:19:13 2025
    On 07/08/2025 20:35, olcott wrote:
    On 8/7/2025 2:14 PM, James Kuyper wrote:
    On 06/08/2025 23:50, olcott wrote:
    ...
    How valuable to computer science would it be to understand
    that the conventional halting problem proofs do not actually
    derive their undecidability conclusion?

    Understanding a falsehood to be true would have a very large
    negative
    value. The validity of the proof of the halting problem isn't
    hard for
    most mathematically minded people to understand - the hard part
    was
    coming up with it.

    Unless you want to understand the notion of the
    recursive simulation non-halting behavior pattern
    please do not reply.

    The recursive nature of your code is irrelevant because it is
    only when it has all unwound and HHH has returned for the last
    time that DD makes its decision. HHH *cannot* report on the
    outermost DD call because the outermost DD's termination status
    cannot be determined until the outermost HHH's return value is known.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Thu Aug 7 23:06:30 2025
    On 07/08/2025 21:54, olcott wrote:
    On 8/7/2025 3:19 PM, Richard Heathfield wrote:

    <snip>

    The recursive nature of your code is irrelevant because it is
    only when it has all unwound and HHH has returned for the last
    time that DD makes its decision.

    Incorrect. When HHH(DD) aborts its simulation

    May I remind you that in comp.lang.c HHH can't even begin its
    simulation because it has no legal way to get at DD's instructions.

    the
    separate process context DD stack is no longer accessible.

    That's nothing to do with C.

    What C can tell us is that (however it obtains its result) if HHH
    reports "DD halts" DD loops, and if HHH reports "DD loops" DD
    halts. So no matter how HHH does its work it gets the answer wrong.

     HHH *cannot* report on the outermost DD call because the
    outermost DD's termination status cannot be determined until
    the outermost HHH's return value is known.


    Counter-factual

    No, sir. Self-evident.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kaz Kylheku@21:1/5 to olcott on Thu Aug 7 22:59:08 2025
    On 2025-08-07, olcott <[email protected]> wrote:
    Three different LLM AI systems figured this all out on
    their own as soon as they were given this notion of a
    simulating termination analyzer:

    Please, just stop the LLM stuff.

    Or at least explore it in areas where you are sure you have iron-clad
    knowledge (i.e. anything but halting).

    LLMs are just "Eliza on steroids" token predictors that predict
    tokens based o large volumes of training texts, and effectively
    interpolate among the data.

    The commercial models are programmed to be sycophantic to the end users
    in order to promote the popularity of AI.

    Crackpots all over the world are using AI to validate their
    cockamamie theories, since humans refuse.

    Language models often spew outright incorrect statements
    ("hallucinations"); you have to be knowledgable in the area you are
    conversing about in order to spot the falsehoods, or otherwise check the output. For instance, if you ask for code, you can tell that it
    doesn't work.

    Quite recently, LLM AI insisted me that a certain configuration change
    in Windows was /not/ responsible for the issue I was trying to
    investigate, and tried to persuade me that leaving the configuration
    as-is was necessary for getting things working. In fact, it /was/
    that configuration; I flipped it, rebooted, and the problem was fixed.

    Currently breaking story: "After using ChatGPT, man swaps his salt for
    sodium bromide?and suffers psychosis"

    https://arstechnica.com/health/2025/08/after-using-chatgpt-man-swaps-his-salt-for-sodium-bromide-and-suffers-psychosis/

    Older: "ChatGPT Made My Husband Think He's God"

    https://medium.com/write-a-catalyst/chatgpt-made-my-husband-think-hes-god-the-ai-apocalypse-destroying-american-families-5f33e4d04a51

    --
    TXR Programming Language: http://nongnu.org/txr
    Cygnal: Cygwin Native Application Library: http://kylheku.com/cygnal
    Mastodon: @[email protected]

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to Kaz Kylheku on Fri Aug 8 00:27:31 2025
    On 07/08/2025 23:59, Kaz Kylheku wrote:
    For instance, if you ask for code, you can tell that it
    doesn't work.

    ...or it does.

    Perfect it isn't, but it's impressive even so.

    You have to be firm, and you have to be clear about what you
    want, but it's surprisingly adept at turning intentions into
    working code.

    On the other hand, it fails disappointingly often to count the
    number of letters in a word.

    Let's see how it copes with olcott's code:

    I am testing the assumption that simulating termination analyzer
    HHH correctly simulates its input until it:
    (a) Detects a non-terminating behavior pattern: abort simulation
    and return 0.
    (b) Simulated input reaches its simulated "return" statement:
    return 1.

    typedef int (*ptr)();
    int HHH(ptr P);

    int DD()
    {
    int Halt_Status = HHH(DD);
    if (Halt_Status)
    HERE: goto HERE;
    return Halt_Status;
    }

    What value should HHH(DD) correctly return?
    What does that tell me about my assumption? --------------------------------------------


    Alright — let’s unpack your test case step-by-step.
    1. Restating your setup

    You have:

    typedef int (*ptr)();
    int HHH(ptr P); // Halting analyser simulator: returns 0 if
    detects nontermination, 1 if detects return

    int DD()
    {
    int Halt_Status = HHH(DD);
    if (Halt_Status)
    HERE: goto HERE;
    return Halt_Status;
    }

    Your rules for HHH:

    Return 0 → You’ve detected that P() will run forever.

    Return 1 → You’ve simulated P() to the point it would return.

    2. Tracing what happens if HHH works per your assumption

    Let’s pretend HHH(DD) really does what you hope — simulate
    exactly until it sees one of your stopping conditions.
    Case A: HHH(DD) returns 1

    In DD():

    if (Halt_Status) // 1 means true
    HERE: goto HERE; // infinite loop

    So if HHH predicts “halts”, the actual execution of DD will
    loop forever.

    That means HHH’s prediction was wrong — it said “halts”, but
    it doesn’t.

    Case B: HHH(DD) returns 0

    Then:

    if (0) // false
    HERE: goto HERE;
    return 0;

    So if HHH predicts “loops forever”, DD() actually terminates
    (returns 0 immediately).

    Again, prediction is wrong — it said “nonterminates”, but it
    halts.

    3. The paradox

    In both cases, whatever HHH returns is falsified by the actual
    behaviour of DD():
    HHH(DD) says... Actual behaviour of DD()
    1 (halts) loops forever
    0 (loops) halts
    This is basically a diagonalisation: DD is constructed to do the
    opposite of what HHH predicts about it. --------------------------------------------

    Full marks to ChatGPT.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 00:44:48 2025
    On 08/08/2025 00:32, olcott wrote:
    On 8/7/2025 5:06 PM, Richard Heathfield wrote:
    On 07/08/2025 21:54, olcott wrote:
    On 8/7/2025 3:19 PM, Richard Heathfield wrote:

    <snip>

    The recursive nature of your code is irrelevant because it is
    only when it has all unwound and HHH has returned for the
    last time that DD makes its decision.

    Incorrect. When HHH(DD) aborts its simulation

    May I remind you that in comp.lang.c HHH can't even begin its
    simulation because it has no legal way to get at DD's
    instructions.


    Yes of course we all know that it is completely
    impossible for any C program to read binary files
    such as the COFF object file named Halt7.obj.

    Clearly not, but HHH doesn't do that. Are you planning a new
    version of HHH that uses fopen and fread to access the object
    file? That would certainly be possible to do in C, but that isn't
    how HHH works, so it's a blind alley.



    the
    separate process context DD stack is no longer accessible.

    That's nothing to do with C.

    What C can tell us is that (however it obtains its result) if
    HHH reports "DD halts" DD loops, and if HHH reports "DD loops"
    DD halts. So no matter how HHH does its work it gets the answer
    wrong.


    I will keep insisting that you understand this
    otherwise you are cheating.

    Cheating? It's not a game. It's logic.


    *recursive simulation non-halting behavior pattern*
    Giving a review of a work without looking at the work
    is dishonest.

    The opinion I gave was on your DD function, the code for which
    you posted, and which I used to explain why your conclusion is false.

    If you don't want to be told why your code doesn't prove what you
    think it proves, don't post it.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 00:31:46 2025
    On 08/08/2025 00:28, olcott wrote:
    On 8/7/2025 5:59 PM, Kaz Kylheku wrote:
    On 2025-08-07, olcott <[email protected]> wrote:
    Three different LLM AI systems figured this all out on
    their own as soon as they were given this notion of a
    simulating termination analyzer:

    Please, just stop the LLM stuff.

    Or at least explore it in areas where you are sure you have
    iron-clad
    knowledge (i.e.  anything but halting).

    LLMs are just "Eliza on steroids" token predictors that predict
    tokens based o large volumes of training texts, and effectively
    interpolate among the data.

    The commercial models are programmed to be sycophantic to the
    end users
    in order to promote the popularity of AI.

    Crackpots all over the world are using AI to validate their
    cockamamie theories, since humans refuse.

    Language models often spew outright incorrect statements
    ("hallucinations"); you have to be knowledgable in the area you
    are
    conversing about in order to spot the falsehoods, or otherwise
    check the
    output. For instance, if you ask for code, you can tell that it
    doesn't work.

    Quite recently, LLM AI insisted me that a certain configuration
    change
    in Windows was /not/ responsible for the issue I was trying to
    investigate, and tried to persuade me that leaving the
    configuration
    as-is was necessary for getting things working. In fact, it /was/
    that configuration; I flipped it, rebooted, and the problem was
    fixed.

    Currently breaking story: "After using ChatGPT, man swaps his
    salt for
    sodium bromide?and suffers psychosis"

    https://arstechnica.com/health/2025/08/after-using-chatgpt-man-swaps-his-salt-for-sodium-bromide-and-suffers-psychosis/

    Older: "ChatGPT Made My Husband Think He's God"

    https://medium.com/write-a-catalyst/chatgpt-made-my-husband-think-hes-god-the-ai-apocalypse-destroying-american-families-5f33e4d04a51


    None-the-less ALL three LLM systems were able to
    correctly figure out the
    *recursive simulation non-halting behavior pattern*
    entirely on the basis of this definition:

    Simulating Termination Analyzer HHH correctly simulates its input
    until:
    (a) Detects a non-terminating behavior pattern: abort simulation
    and return 0.
    (b) Simulated input reaches its simulated "return" statement:
    return 1.


    Having announced that Turing was right, ChatGPT went on to say:

    your DD is essentially the C embodiment of the classic “ask the
    oracle and then do the opposite” trick.

    If you keep going down this rabbit hole, you can also phrase your
    conclusion as:

    Any HHH that halts with a definite answer for all inputs is
    either incorrect or non-total.

    That’s the Halting Problem in a crisp, runnable form.

    Do you want me to also walk through what happens if HHH tries to
    be “clever” and avoid the contradiction by simulating DD lazily
    or partially?
    You said:
    yes, please
    ChatGPT said:

    Alright — let’s walk through what happens if HHH tries to “get
    clever” and avoid the paradox.

    Your current assumption was that HHH:

    Simulates its input program step-by-step.

    Stops when either:

    It recognises a looping pattern → returns 0
    (“non-terminates”).

    It reaches a return → returns 1 (“terminates”).

    Now we’ll try possible evasions.
    1. “Lazy” simulation

    Idea: HHH doesn’t simulate fully — maybe it only runs P() for n
    steps, then guesses.

    Problem: Any finite n can be exceeded by some program’s
    pre-loop behaviour.

    For DD, if you guess after n steps, you can still be wrong:

    Guess “halts” → actual runs forever.

    Guess “loops” → actual halts.

    This is just giving the wrong answer sooner.

    Result: Still fails — you’re just getting the contradiction quicker.
    2. Pattern detection without full simulation

    Idea: Instead of simulating to completion, detect certain code
    shapes or loop forms and decide from that.

    Works for some programs (e.g., obvious while(1) with no breaks).

    But DD hides its loop behind a call to HHH about itself.

    You can’t see the final control flow without essentially
    solving the halting problem on HHH + DD — which is exactly the
    unsolvable bit.

    Result: Pattern matching can be tricked by opaque cases, so
    you’ll miss some and fail correctness.
    3. Refusal to decide (“third option”)

    Idea: If HHH can’t decide, it says “I don’t know” instead of 0 or 1.

    This can avoid the contradiction for specific cases like DD.

    But it means HHH is partial — it doesn’t give an answer for
    all inputs.

    That’s exactly the escape route allowed by computability
    theory: you can be correct on what you do answer, but you can’t
    be correct and complete for all possible inputs.

    Result: No contradiction, but you’ve abandoned totality.
    4. Make HHH non-terminating for DD

    Idea: When HHH is given a self-referential input like DD, it just
    loops forever instead of returning.

    That “works” in the sense of not returning a wrong answer —
    but it also means HHH isn’t total.

    This is actually the only way to make a “correct” halting
    analyser in the absolute sense: you let it run forever when you
    can’t decide.

    Result: Correctness preserved, totality lost.
    Key takeaway

    The clever tricks all boil down to a trade-off:
    Strategy Correct? Answers for all inputs?
    Naïve simulation ❌ ✅
    Guess after timeout ❌ ✅
    Pattern detection only ❌ ✅
    “I don’t know” option ✅ ❌
    Loop forever on hard cases ✅ ❌
    You can build a partially correct HHH that never lies, but it
    must fail to return for some inputs — and DD will be one of them.


    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 07:02:14 2025
    On 08/08/2025 01:11, olcott wrote:
    On 8/7/2025 6:44 PM, Richard Heathfield wrote:
    On 08/08/2025 00:32, olcott wrote:
    On 8/7/2025 5:06 PM, Richard Heathfield wrote:
    On 07/08/2025 21:54, olcott wrote:
    On 8/7/2025 3:19 PM, Richard Heathfield wrote:

    <snip>

    The recursive nature of your code is irrelevant because it
    is only when it has all unwound and HHH has returned for
    the last time that DD makes its decision.

    Incorrect. When HHH(DD) aborts its simulation

    May I remind you that in comp.lang.c HHH can't even begin its
    simulation because it has no legal way to get at DD's
    instructions.


    Yes of course we all know that it is completely
    impossible for any C program to read binary files
    such as the COFF object file named Halt7.obj.

    Clearly not, but HHH doesn't do that. Are you planning a new
    version of HHH that uses fopen and fread to access the object
    file? That would certainly be possible to do in C, but that
    isn't how HHH works, so it's a blind alley.


    https://github.com/plolcott/x86utm/blob/master/include/Read_COFF_Object.h

    That isn't the HHH code, and it isn't how HHH works, so it's a
    blind alley.

    The opinion I gave was on your DD function, the code for which
    you posted, and which I used to explain why your conclusion is
    false.

    If you don't want to be told why your code doesn't prove what
    you think it proves, don't post it.


    recursive simulation non-halting behavior pattern

    ...is irrelevant to the question of whether HHH does what you
    imagine it does. Tis tis tis tis tis does not constitute a proof.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kaz Kylheku@21:1/5 to Richard Heathfield on Fri Aug 8 05:59:52 2025
    On 2025-08-07, Richard Heathfield <[email protected]> wrote:
    Clearly not, but HHH doesn't do that. Are you planning a new
    version of HHH that uses fopen and fread to access the object
    file? That would certainly be possible to do in C, but that isn't
    how HHH works, so it's a blind alley.

    A numbrer of years ago, Peter had a working system (according to
    plausible evidence from detailed execution traces) consisting of a
    modified x86 simulator, for which he compiled C code. It was able to do
    the introspection and nested simulations you are hinting at.

    The only problem is that he believed that to be proving things that it
    wasn't proving.

    I don't think he's given anyone the materials to run for themselves,
    but I analyzed his execution traces in some detail.

    If my memory serves me, he endowed the simulator (existing work
    from someone else) with logic that tried to detect the looping
    behavior of the halting test case (the code that tries to "behave
    opposite" to the halting decision).

    I'm afraid this line of discussion is blind alley in a different sense.

    Peter Olcott "been there and done that" in terms of getting up to the
    elbows in a CPU simulating tech stack and trying to make a synthetic CPU
    that runs compiled C code and all that.

    I remember that one of the issues with this solutions (as far as I
    was able to guess from his execution traces) was that he hacked
    the x86 simulator to detect, in the nested simulated situation,
    when it was simulating its own code. And that was part of where
    the reasonign went astray; he thought you can pull a ruse like that,
    and still be proving something about halting.

    What happens in the halting proof is that the rebellious program
    that behaves opposite has access to a halting decision about itself,
    and that requires it to /embed/ the halting decider. I.e. given
    a halting we can build a program /around/ it which reacts to
    its decision by infinitely looping when the decision says "halts",
    or terminating when it says "doesn't halt".

    Olcott's system, however, had built-in detection for when the
    instruction pointer entered into the halting decider (I think simply by comparing the addresses), and it pulled some invalid shennanigans there,
    gaming the control flow. I remember seeing impossible events in the instruction-by-instruction execution traces, like the instruction
    pointer jumping somwhere without having executed a branch.

    It is all actually admirable hacking effort, but for the muddled
    thinking and wrong conclusions.

    At that point I realized that Peter's conviction is likely unbreakable.
    Or, if there is a way, it has to be via the fundamentals. Getting him to
    accept the diagnonalizaton argument, and the concepts of pure functions
    and all the rest of it.

    --
    TXR Programming Language: http://nongnu.org/txr
    Cygnal: Cygwin Native Application Library: http://kylheku.com/cygnal
    Mastodon: @[email protected]

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to Kaz Kylheku on Fri Aug 8 07:46:51 2025
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable.

    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Janis Papanagnou@21:1/5 to Richard Heathfield on Fri Aug 8 16:10:40 2025
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable.

    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    (This pathological thread, revived and re-triggered by the unwary,
    now lasts for years without any new insights; it's just pointless reiterations.)

    No party will convince the other; this is quite obvious since long.
    I think *this* is the relevant "key insight". - Act accordingly![*]

    Janis

    [*] Just a suggestion [for folks who try to avoid useless frays].

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 17:23:38 2025
    On 08/08/2025 15:09, olcott wrote:
    On 8/8/2025 12:59 AM, Kaz Kylheku wrote:

    <snip>


    DD correctly simulated by HHH cannot possibly reach
    the "do the opposite" code because it remains stuck
    in recursive simulation.

    So what you're saying is that HHH cannot correctly simulate DD.

    <snip>

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to Janis Papanagnou on Fri Aug 8 17:24:28 2025
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable.

    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 19:12:43 2025
    On 08/08/2025 15:27, olcott wrote:

    <snip>


    It is a verified fact that DD correctly simulated by HHH
    cannot possibly reach its own simulated "return" statement
    final halt state.

    Then it's not a correct simulation.

    It is a verified fact that HHH(DD) correctly detects this.
    I am attesting to the fact that Claude AI did correctly
    determine why HHH(DD)==0 is correct.

    If that's correct, HHH's judgement is that DD doesn't halt.
    Having received that report from HHH, DD promptly halts,
    invalidating HHH's judgement, so it's not correct after all. HHH
    fails in its mission as a halt decider.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 19:29:30 2025
    On 08/08/2025 17:40, olcott wrote:
    Then it can be directly seen that DD correctly simulated
    by HHH cannot possibly reach its own simulated "return"
    statement final halt state.

    And therefore you have derived a contradiction which demonstrates
    that your reasoning contains a false assumption, namely that DD
    is correctly simulated by HHH.

    QED.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Mike Terry@21:1/5 to Kaz Kylheku on Fri Aug 8 20:51:27 2025
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    On 2025-08-07, Richard Heathfield <[email protected]> wrote:
    Clearly not, but HHH doesn't do that. Are you planning a new
    version of HHH that uses fopen and fread to access the object
    file? That would certainly be possible to do in C, but that isn't
    how HHH works, so it's a blind alley.

    A numbrer of years ago, Peter had a working system (according to
    plausible evidence from detailed execution traces) consisting of a
    modified x86 simulator, for which he compiled C code. It was able to do
    the introspection and nested simulations you are hinting at.

    The only problem is that he believed that to be proving things that it
    wasn't proving.

    I don't think he's given anyone the materials to run for themselves,
    but I analyzed his execution traces in some detail.

    From time to time he publishes a link to all the source code. I think he posted a link today?
    It's easy enough to compile if you're using MSVC (on Windows). It shouldn't be hard to port to
    other environments, but I guess that assumes the programmer already understands the code - PO
    doesn't explain anything much. So in practice you'd need to play using the MSVC toolset to get the
    basic understanding before starting any port! [ok, so it won't be so easy...]


    If my memory serves me, he endowed the simulator (existing work
    from someone else) with logic that tried to detect the looping
    behavior of the halting test case (the code that tries to "behave
    opposite" to the halting decision).

    I'm afraid this line of discussion is blind alley in a different sense.

    Peter Olcott "been there and done that" in terms of getting up to the
    elbows in a CPU simulating tech stack and trying to make a synthetic CPU
    that runs compiled C code and all that.

    I remember that one of the issues with this solutions (as far as I
    was able to guess from his execution traces) was that he hacked
    the x86 simulator to detect, in the nested simulated situation,
    when it was simulating its own code. And that was part of where
    the reasonign went astray; he thought you can pull a ruse like that,
    and still be proving something about halting.

    Yes, there are a couple of problems here:
    - PO uses global variables to communicate across all emulation levels [his global trace table], and
    for emulated code to recognise when it is being emulated and behave differently [his Root variable].
    This is obviously a No-No.
    - PO has coded a decider H which recognised the emulated computation calling H by comparing the
    emulated call with its own address.

    The first problem breaks the design for his "emulation", since emulated code by definition MUST
    match the behaviour of the code being emulated. This problem could be fixed, but not by PO who
    can't work out how to fix it. He says he doesn't have a thousand years to make it work "properly"!
    [It took me about an hour to fix in my play code, once I'd worked out what it was doing...]

    - The second problem [using H's own code address] does present a sort of puzzle for how this code
    relates to TMs, since TM states don't have "addresses" that they can access. OTOH this is not
    really the reason PO's argument is nonsense, so I don't see this as a problem worth focussing on.
    If PO's decider H actually contradicted anything in the Linz proof, we might start looking for
    deeper reasons, but given it /doesn't/, arging about fiddly technical details is not worth the
    bother IMO.

    Note PO has a different decider implementation HHH which does not use its own address, instead
    watching for repeated addresses within nested emulations.


    What happens in the halting proof is that the rebellious program
    that behaves opposite has access to a halting decision about itself,
    and that requires it to /embed/ the halting decider. I.e. given
    a halting we can build a program /around/ it which reacts to
    its decision by infinitely looping when the decision says "halts",
    or terminating when it says "doesn't halt".

    Olcott's system, however, had built-in detection for when the
    instruction pointer entered into the halting decider (I think simply by comparing the addresses), and it pulled some invalid shennanigans there, gaming the control flow.

    Probably that's the H code address problem (see above). Being charitable to PO, that's not what
    leads to his incorrect conclusions (see below).

    I remember seeing impossible events in the
    instruction-by-instruction execution traces, like the instruction
    pointer jumping somwhere without having executed a branch.

    That could be a red herring. PO filters the trace output he presents to only show instructions
    within DDD() function [the test program input that decider is analysing], not within the (logically)
    embedded decider HHH. So when DDD() calls HHH it /looks/ like the IP has jumped not to HHH, but to
    DDD, but that is not really the case - more that PO doesn't bother to explain what is actually
    happening because he claims that just confuses people! [Yeah right - pretty much every poster who
    comes along is confused by why the instruction has jumped to DDD when it should have gone to HHH,
    but PO doesn't count that as confusing people, lol.] And of course, the filtering /is/ a problem if
    you reasonably expect to trace through and inspect all the code (HHH/DDD/+subroutines) to satisfy
    your concerns.


    It is all actually admirable hacking effort, but for the muddled
    thinking and wrong conclusions.

    Right. The Linz proof shows that if DDD is constructed by wrapping decider HHH as you've described,
    and if HHH decides DDD never halts [like PO's HHH], then in fact DDD must halt [contradiction! so
    HHH isn't in fact a halt decider]. So... after PO's years of work creating his x86utm.exe (runtime
    environment) and HHH/DDD (halt decider/input code) what was the result? His HHH decides DDD never
    halts, and in fact DDD is quickly demonstrated to actually halt, along with all the traces backing
    up these observations. Ie. EXACTLY IN LINE WITH LINZ proof.

    Lol, 2 years development work totally wasted... Everyone would have happily agreed in advance that
    this is the expected behaviour for the code, provided there was no "cheating" tricks, so developing
    a complicated program to "prove" that achieved nothing.


    At that point I realized that Peter's conviction is likely unbreakable.
    Or, if there is a way, it has to be via the fundamentals. Getting him to accept the diagnonalizaton argument, and the concepts of pure functions
    and all the rest of it.

    I'm afraid I think concentrating on concepts is a hopeless approach, because PO simply /can't/ do
    any of that, which I put down to alternative brain wiring: PO's brain is unable to handle abstract
    concepts and logical reasoning in the way other people can effortlessly.

    Similarly, using "logical reasoning" to try to convince him is a waste of time - I think he
    literally doesn't register the logic. [My imagining of PO's thinking: "Hey, what's the point of
    what you just said? I'm not sure what you were getting at, but it seems you're apparently trying to
    tell me that something I said isn't true. I can't let that go in case future readers think you
    might have a point - but I can't meaningfully address whatever it was you were saying, so I'll
    ignore all that and just repeat what I've said before until you see the light."]

    Other approaches? For a while I thought "concrete demonstrations" might stand a chance, but
    experience proves otherwise! Consider - PO has already faced down the concrete demonstration that
    DDD halts, whereas HHH(DDD) decides DDD never halts. In effect he has just carried on as before
    saying More Words devoid of logical content, ending "...so HHH is /right/ to say DDD never halts
    even though DDD halts. You have to be a genius to understand the truth of my words, which is why
    you don't agree etc. etc."

    So /explaining/ doesn't work, /reasoning/ doesn't work, and /showing/ him doesn't work. Is there
    anything else left? :)

    The better news is that PO is not doing any harm - nobody is convinced by his claims, and he cannot
    affect your children's education or anything in your life, unless you /choose/ to engage with him.
    So HE DOESN'T NEED TO BE "DEFEATED". We can just let him get on with his master plan [step 1:
    publish paper in respectable peer-reviewed journal] Yes, at some point he'll die believing he's a
    genius surrounded by idiots, and that he's proved all his claims over and over, and even some
    chatbots agree. That's all perfectly OK. The world will carry on exactly as before. [except
    comp.theory will soon die off :( ]


    Mike.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 22:00:58 2025
    On 08/08/2025 20:35, olcott wrote:
    On 8/8/2025 1:29 PM, Richard Heathfield wrote:
    On 08/08/2025 17:40, olcott wrote:
    Then it can be directly seen that DD correctly simulated
    by HHH cannot possibly reach its own simulated "return"
    statement final halt state.

    And therefore you have derived a contradiction which
    demonstrates that your reasoning contains a false assumption,
    namely that DD is correctly simulated by HHH.

    QED.


    HHH(DD) only verifies whether or not DD emulated by HHH
    according to the semantics x86 language reaches its own
    emulated "ret" instruction.

    How HHH decides is of no consequence. All that matters is /what/
    it decides, because what it decides is wrong.

    It is flatly incorrect to disagree with the semantics
    of the x86 language.

    It is flatly incorrect to think that the semantics of x86 machine
    code matter a damn. For all I care HHH could get its answer by
    waving a wand and saying abracadabra. It's still the wrong answer.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 21:57:40 2025
    On 08/08/2025 20:29, olcott wrote:

    <snip>

    HHH(DD) only verifies whether or not DD emulated by HHH
    according to the semantics x86 language reaches its own
    emulated "ret" instruction.

    How HHH reaches its decision about whether DD halts is
    irrelevant. What matters is that it reaches a decision that is
    guaranteed to be wrong.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Fri Aug 8 23:45:51 2025
    On 08/08/2025 23:22, olcott wrote:
    On 8/8/2025 3:57 PM, Richard Heathfield wrote:
    On 08/08/2025 20:29, olcott wrote:

    <snip>

    HHH(DD) only verifies whether or not DD emulated by HHH
    according to the semantics x86 language reaches its own
    emulated "ret" instruction.

    How HHH reaches its decision about whether DD halts is
    irrelevant. What matters is that it reaches a decision that is
    guaranteed to be wrong.


    It only seems that way on the basis of a misconception.

    The misconception is yours.

    This is not a mystical cloud of unknowing; it's simple logic.

    You can't begin to understand that it is a misconception

    It isn't a misconception.

    until after you understand this:

    Claude AI proved why HHH(DD)==0 is correct in terms that
    any expert C programmer can understand. https://claude.ai/share/da9e56ba-f4e9-45ee-9f2c-dc5ffe10f00c

    Complete bollocks from the AI:

    "This is a classic example of a halting problem scenario where
    the analyzer correctly identifies that the input represents
    non-terminating behavior, even though the actual execution might
    terminate"

    It can't bear to disappoint you, so it finds some wording that
    will placate you, but it's self-contradictory nonsense.


    How is it self-contradictory? Well, let's take "Answer: HHH(DD)
    should correctly return 0." like the AI says.

    HHH returns 0, so in DD the if()_fails, and DD returns after all.

    HHH got it wrong.

    Which is exactly what you'd expect, because DD is designed to
    make HHH fail.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Sat Aug 9 00:07:58 2025
    On 08/08/2025 23:31, olcott wrote:
    On 8/8/2025 4:00 PM, Richard Heathfield wrote:
    On 08/08/2025 20:35, olcott wrote:
    On 8/8/2025 1:29 PM, Richard Heathfield wrote:
    On 08/08/2025 17:40, olcott wrote:
    Then it can be directly seen that DD correctly simulated
    by HHH cannot possibly reach its own simulated "return"
    statement final halt state.

    And therefore you have derived a contradiction which
    demonstrates that your reasoning contains a false assumption,
    namely that DD is correctly simulated by HHH.

    QED.


    HHH(DD) only verifies whether or not DD emulated by HHH
    according to the semantics x86 language reaches its own
    emulated "ret" instruction.

    How HHH decides is of no consequence. All that matters is
    /what/ it decides, because what it decides is wrong.


    It is wrong to assume that any computable function
    must report on the behavior of anything besides
    the behavior that its actual input actually specifies.

    For a start, you're assuming that HHH is computable.

    Secondly, all I'm assuming is that HHH is giving its best stab at
    deciding whether DD halts.

    That's its job. If it isn't at least trying to do that job, you
    don't have anything.

    DD emulated by HHH according to the semantics x86
    language conclusively proves the behavior that the
    input to HHH(DD) actually specifies.

    Yawn. Does DD halt or doesn't it? If HHH can't find that out by
    emulation, then emulation is not up to the job.


    It is flatly incorrect to disagree with the semantics
    of the x86 language.

    It is flatly incorrect to think that the semantics of x86
    machine code matter a damn. For all I care HHH could get its
    answer by waving a wand and saying abracadabra. It's still the
    wrong answer.


    Without the exactingly precise (state transition diagram)
    degree of precision that the x86 language specifies we could
    go around and around forever on the basis of incorrect measures
    of correct simulation.

    Well, I'll leave you to get on with that, but the exacting
    whatever is all sugar on your fluff cake.

    ALL THAT MATTERS is the answer. If you get that wrong, you have
    nothing.

    And you do.


    If you have the opinion that 3 > 5 then you are just wrong.

    That is not an opinion I hold.

    If you have the opinion about correct simulation

    I don't give a damn whether the simulation is correct. You do, of
    course, but that's not a headache I need to share. All I need is
    a number out of HHH - a 0, or a 1.

    And whichever I get, it turns out to be the wrong number.


    and there
    is no absolute measure then there is no closure. The x86
    language is *the absolute measure of correct simulation*

    I don't care how the simulation works. What number does it spit
    out, 0 or 1? Give me that number, and I'll show you it's wrong.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to olcott on Sat Aug 9 00:17:50 2025
    On 09/08/2025 00:00, olcott wrote:

    <snip>


    You must carefully study the whole page until
    you have 100% complete understanding of the
    *recursive simulation non-halting behavior pattern*

    Wrong.

    All we need is the code for DD.

    DD correctly simulated by HHH never reaches the "if".

    So what?

    HHH must report to its caller. Its caller is DD. When it
    eventually returns to DD, DD continues executing.

    HHH got it wrong.

    Which is exactly what you'd expect, because DD is designed to
    make HHH fail.


    Termination analyzers in general are only accountable
    for the actual behavior that their input actually specifies.

    A termination analyser that doesn't determine whether DD
    terminates is not a termination analyser worth the bytes it occupies.


    HHH(DD)==0 is correct on the basis of the behavior that
    the actual input actually specifies.

    So HHH returns 0, so DD returns, making HHH's report incorrect.


    That computer scientists did not notice this for 89 years
    is not my mistake.

    Your mistake is in misunderstanding a simple proof.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to Keith Thompson on Sat Aug 9 00:28:09 2025
    On 09/08/2025 00:21, Keith Thompson wrote:
    Richard Heathfield <[email protected]> writes:
    On 08/08/2025 23:31, olcott wrote:
    [...]

    I observe that comp.lang.c is now dominated by a discussions of
    olcott's claims, and that those discussions now have little or not
    C content.

    This is not addressed to olcott. He has demonstrated that he will
    ignore topicality complaints, and I don't see anything he posts here.


    But you know I'll take you seriously.

    And of course I will.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Kaz Kylheku@21:1/5 to olcott on Sat Aug 9 03:06:12 2025
    On 2025-08-08, olcott <[email protected]> wrote:
    On 8/8/2025 12:59 AM, Kaz Kylheku wrote:
    I don't think he's given anyone the materials to run for themselves,
    but I analyzed his execution traces in some detail.


    *The whole system*
    https://github.com/plolcott/x86utm/

    Ah, you published it! I think that is great.

    Regardless of it not being able to defeat halting, it can be regarded as artistic object, capturing how far someone was willing to go who is
    convinced that an unassilable theorem is wrong.

    Trying to defeat academia with x86 programming has a kind of
    Don Quixote vibe to it, if we consider hacking on code to be
    a kind of chivalry.

    CS academia has lost a lot of that. Decades ago, you could develop a
    system, and publish about it. Much less so today.

    When papers mention some code these days, it is nowhere to be found.
    "Believe us when we say that we coded all this and these graphs
    and tables are real ...".

    --
    TXR Programming Language: http://nongnu.org/txr
    Cygnal: Cygwin Native Application Library: http://kylheku.com/cygnal
    Mastodon: @[email protected]

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Janis Papanagnou@21:1/5 to Richard Heathfield on Sat Aug 9 15:36:29 2025
    On 08.08.2025 18:24, Richard Heathfield wrote:
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable. >>>
    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    About five decades ago. - Does that explain anything WRT this useless
    thread? - Are you (still?) chewing gum?

    Anyway. - I got rid of Olcott's band worms of OT reiterations only by killfiling his name. - Now there's a zombie band worm of reiterations
    from another single person... *sigh* - So be it. Have fun!

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From olcott@21:1/5 to Janis Papanagnou on Sat Aug 9 09:13:21 2025
    On 8/9/2025 8:36 AM, Janis Papanagnou wrote:
    On 08.08.2025 18:24, Richard Heathfield wrote:
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable. >>>>
    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    About five decades ago. - Does that explain anything WRT this useless
    thread? - Are you (still?) chewing gum?

    Anyway. - I got rid of Olcott's band worms of OT reiterations only by killfiling his name. - Now there's a zombie band worm of reiterations
    from another single person... *sigh* - So be it. Have fun!


    typedef int (*ptr)();
    int HHH(ptr P);

    int DD()
    {
    int Halt_Status = HHH(DD);
    if (Halt_Status)
    HERE: goto HERE;
    return Halt_Status;
    }

    If people would not spend hundreds of hours coming
    up with ill-formed rebuttals and instead spent five
    minutes actually understanding that

    DD correctly simulated by HHH cannot possibly reach
    its own simulated "if" statement or its own simulated
    "return" statement because DD calls HHH(DD) in recursive
    simulation

    I would have have been done with the group years ago.

    I am not here for any comp.theory aspects of the above
    I am only here for the semantics of the C programming
    language.

    It was Mike that delved into the deep comp.theory aspects
    of this on this group. I am only here for the C aspects.



    --
    Copyright 2024 Olcott

    "Talent hits a target no one else can hit;
    Genius hits a target no one else can see."
    Arthur Schopenhauer

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Richard Heathfield@21:1/5 to Janis Papanagnou on Sat Aug 9 15:13:38 2025
    On 09/08/2025 14:36, Janis Papanagnou wrote:
    On 08.08.2025 18:24, Richard Heathfield wrote:
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable. >>>>
    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    About five decades ago. - Does that explain anything WRT this useless
    thread? - Are you (still?) chewing gum?

    Anyway. - I got rid of Olcott's band worms of OT reiterations only by killfiling his name. - Now there's a zombie band worm of reiterations
    from another single person... *sigh* - So be it. Have fun!

    Yeah, likewise, but the crossposts might take a while to settle down.

    --
    Richard Heathfield
    Email: rjh at cpax dot org dot uk
    "Usenet is a strange place" - dmr 29 July 1999
    Sig line 4 vacant - apply within

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Mike Terry@21:1/5 to Janis Papanagnou on Sat Aug 9 16:16:02 2025
    On 09/08/2025 14:36, Janis Papanagnou wrote:
    On 08.08.2025 18:24, Richard Heathfield wrote:
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely unbreakable. >>>>
    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    About five decades ago. - Does that explain anything WRT this useless
    thread? - Are you (still?) chewing gum?

    Anyway. - I got rid of Olcott's band worms of OT reiterations only by killfiling his name. - Now there's a zombie band worm of reiterations
    from another single person... *sigh* - So be it. Have fun!


    The newsreaders I've used all support a "ignore subthread" action. If you can use that, you won't
    see PO's posts or any replies (unless you use an option to "view ignored posts").


    Mike.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Janis Papanagnou@21:1/5 to Mike Terry on Sat Aug 9 18:05:49 2025
    On 09.08.2025 17:16, Mike Terry wrote:
    On 09/08/2025 14:36, Janis Papanagnou wrote:
    On 08.08.2025 18:24, Richard Heathfield wrote:
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely
    unbreakable.

    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    About five decades ago. - Does that explain anything WRT this useless
    thread? - Are you (still?) chewing gum?

    Anyway. - I got rid of Olcott's band worms of OT reiterations only by
    killfiling his name. - Now there's a zombie band worm of reiterations
    from another single person... *sigh* - So be it. Have fun!


    The newsreaders I've used all support a "ignore subthread" action. If
    you can use that, you won't see PO's posts or any replies (unless you
    use an option to "view ignored posts").

    I see also in this thread some responses that I find interesting, so
    filtering the subthread is not exactly what I am looking for. Also,
    he regularly started new independent threads with the same contents
    or slightly differing subject lines. - Anyway thanks for your concern.

    What I am looking for is to filter posts from pathological characters.

    As I've seen, sadly, some offensive pathological people deliberately
    decide to bypass personal filter settings by changing nym oder address.

    Olcott (olcott <[email protected]>) has just spent effort to do this,
    it seems.

    We all know that this person is able to understand technical issues,
    yet he has obviously some problems with logical issues. This behavior
    is another indication that he either might logically not understand
    why we have filters (or again just reveals his pathological behavior,
    as he obviously tries to drag me into his lunacy).

    Janis



    Mike.


    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Mike Terry@21:1/5 to Janis Papanagnou on Sat Aug 9 18:00:55 2025
    On 09/08/2025 17:05, Janis Papanagnou wrote:
    On 09.08.2025 17:16, Mike Terry wrote:
    On 09/08/2025 14:36, Janis Papanagnou wrote:
    On 08.08.2025 18:24, Richard Heathfield wrote:
    On 08/08/2025 15:10, Janis Papanagnou wrote:
    On 08.08.2025 08:46, Richard Heathfield wrote:
    On 08/08/2025 06:59, Kaz Kylheku wrote:
    At that point I realized that Peter's conviction is likely
    unbreakable.

    Yes, I think that's the key insight with which anyone should arm
    themselves before launching into the fray.

    But then why continuing to feed him?! - Why engage in the fray in
    the first place?!

    Ever chewed gum?

    About five decades ago. - Does that explain anything WRT this useless
    thread? - Are you (still?) chewing gum?

    Anyway. - I got rid of Olcott's band worms of OT reiterations only by
    killfiling his name. - Now there's a zombie band worm of reiterations
    from another single person... *sigh* - So be it. Have fun!


    The newsreaders I've used all support a "ignore subthread" action. If
    you can use that, you won't see PO's posts or any replies (unless you
    use an option to "view ignored posts").

    I see also in this thread some responses that I find interesting, so filtering the subthread is not exactly what I am looking for. Also,
    he regularly started new independent threads with the same contents
    or slightly differing subject lines. - Anyway thanks for your concern.

    What I am looking for is to filter posts from pathological characters.

    As I've seen, sadly, some offensive pathological people deliberately
    decide to bypass personal filter settings by changing nym oder address.

    Olcott (olcott <[email protected]>) has just spent effort to do this,
    it seems.

    While PO is not without personal faults(!) I wouldn't class him as a nym-shifters. I think he has a
    couple of posting accounts: a GigaNews account and an Eternal September account. I'm pretty sure
    he's not actively trying to evade spam filters... Probably it's just changing newsreaders/servers/reinstalling stuff etc..

    Mike.


    We all know that this person is able to understand technical issues,
    yet he has obviously some problems with logical issues. This behavior
    is another indication that he either might logically not understand
    why we have filters (or again just reveals his pathological behavior,
    as he obviously tries to drag me into his lunacy).

    Janis



    Mike.



    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Mike Terry@21:1/5 to olcott on Sat Aug 9 19:56:01 2025
    On 09/08/2025 18:05, olcott wrote:
    On 8/9/2025 12:00 PM, Mike Terry wrote:
    On 09/08/2025 17:05, Janis Papanagnou wrote:>>
    I see also in this thread some responses that I find interesting, so
    filtering the subthread is not exactly what I am looking for. Also,
    he regularly started new independent threads with the same contents
    or slightly differing subject lines. - Anyway thanks for your concern.

    What I am looking for is to filter posts from pathological characters.

    As I've seen, sadly, some offensive pathological people deliberately
    decide to bypass personal filter settings by changing nym oder address.

    Olcott (olcott <[email protected]>) has just spent effort to do this,
    it seems.

    While PO is not without personal faults(!) I wouldn't class him as a nym-shifters.� I think he has
    a couple of posting accounts: a GigaNews account and an Eternal September account.� I'm pretty
    sure he's not actively trying to evade spam filters...� Probably it's just changing
    newsreaders/servers/reinstalling stuff etc..

    Mike.


    I wanted to make sure that my one reply was seen
    by its author. I have switched back to my usual
    provider for everything else.

    ok I take it back. That's nym-shifting. :( It's actually in-character for you, but I'd just
    never spotted you doing it before...

    Mike.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)