Rocksolid Light

Welcome to novaBBS (click a section below)

mail  files  register  newsreader  groups  login

Message-ID:  

AUTHOR FvwmAuto just appeared one day, nobody knows how. -- FvwmAuto(1x)


computers / comp.theory / Re: Concise refutation of halting problem proofs V62 [ self-evident ]

Re: Concise refutation of halting problem proofs V62 [ self-evident ]

<2eCdnSMfgqk3zZH_nZ2dnUU7-XGdnZ2d@giganews.com>

  copy mid

https://www.novabbs.com/computers/article-flat.php?id=26594&group=comp.theory#26594

  copy link   Newsgroups: comp.theory
Path: i2pn2.org!i2pn.org!weretis.net!feeder6.news.weretis.net!news.misty.com!border2.nntp.dca1.giganews.com!nntp.giganews.com!buffer2.nntp.dca1.giganews.com!news.giganews.com.POSTED!not-for-mail
NNTP-Posting-Date: Tue, 15 Feb 2022 19:17:30 -0600
Date: Tue, 15 Feb 2022 19:17:29 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
Thunderbird/91.6.0
Subject: Re: Concise refutation of halting problem proofs V62 [ self-evident ]
Content-Language: en-US
Newsgroups: comp.theory
References: <vc-dndgn0rt7amL8nZ2dnUU7-LXNnZ2d@giganews.com>
<exZNJ.98392$Rza5.65567@fx47.iad>
<aoCdnfcSnKUy-5X_nZ2dnUU7-aXNnZ2d@giganews.com>
<xXZNJ.38446$Wdl5.5537@fx44.iad>
<ztCdnbKpLJC6AJX_nZ2dnUU7-WednZ2d@giganews.com>
<w96OJ.76395$SeK9.206@fx97.iad>
<1rCdnRW4psaehZT_nZ2dnUU7-VvNnZ2d@giganews.com>
<l4bOJ.5973$kuda.550@fx12.iad>
<8KSdnRooMaz5wpT_nZ2dnUU7-IPNnZ2d@giganews.com>
<DodOJ.37047$Lbb6.31741@fx45.iad>
<aP6dndDBnacS-pT_nZ2dnUU7-anNnZ2d@giganews.com>
<4veOJ.35906$Y1A7.8248@fx43.iad>
<As2dnSSEINq06pT_nZ2dnUU7-eednZ2d@giganews.com>
<wVeOJ.35102$41E7.19078@fx37.iad>
<A-CdnSMdYfQZHZT_nZ2dnUU7-KudnZ2d@giganews.com>
<hmfOJ.40926$Wdl5.7730@fx44.iad>
<L8qdnc1PnZ_-EpT_nZ2dnUU7-RWdnZ2d@giganews.com>
<%ihOJ.13496$GjY3.10711@fx01.iad>
<ZcGdnXE1gbY6PJT_nZ2dnUU7-KudnZ2d@giganews.com>
<geiOJ.14278$jwf9.6136@fx24.iad>
<BtednRZDAaNQUZT_nZ2dnUU7-VHNnZ2d@giganews.com>
<ClrOJ.19559$dln7.7346@fx03.iad> <suglil$dkp$1@dont-email.me>
<9SWOJ.25849$OT%7.13821@fx07.iad>
From: NoO...@NoWhere.com (olcott)
In-Reply-To: <9SWOJ.25849$OT%7.13821@fx07.iad>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Message-ID: <2eCdnSMfgqk3zZH_nZ2dnUU7-XGdnZ2d@giganews.com>
Lines: 466
X-Usenet-Provider: http://www.giganews.com
X-Trace: sv3-F7poN51J7MEG/FTpZDXmmoAUgLe0WL0asboCoqFN5EGDpv+8bum+jENqcgW8smTpQwjNbj3HLshYwli!NjGRBOLciR2pGc2B8r/CxrkBXzdGfXINTfISnj3+mnmkRQ9IDsbsnkjYd5QhyMRIFMeaWINxmaHr
X-Complaints-To: abuse@giganews.com
X-DMCA-Notifications: http://www.giganews.com/info/dmca.html
X-Abuse-and-DMCA-Info: Please be sure to forward a copy of ALL headers
X-Abuse-and-DMCA-Info: Otherwise we will be unable to process your complaint properly
X-Postfilter: 1.3.40
X-Original-Bytes: 27448
 by: olcott - Wed, 16 Feb 2022 01:17 UTC

On 2/15/2022 5:40 PM, Richard Damon wrote:
>
> On 2/15/22 11:49 AM, olcott wrote:
>> On 2/14/2022 5:49 AM, Richard Damon wrote:
>>> On 2/13/22 10:30 PM, olcott wrote:
>>>> On 2/13/2022 7:27 PM, Richard Damon wrote:
>>>>> On 2/13/22 7:26 PM, olcott wrote:
>>>>>> On 2/13/2022 6:24 PM, Richard Damon wrote:
>>>>>>> On 2/13/22 6:08 PM, olcott wrote:
>>>>>>>> On 2/13/2022 4:11 PM, Richard Damon wrote:
>>>>>>>>> On 2/13/22 5:04 PM, olcott wrote:
>>>>>>>>>> On 2/13/2022 3:40 PM, Richard Damon wrote:
>>>>>>>>>>> On 2/13/22 4:24 PM, olcott wrote:
>>>>>>>>>>>> On 2/13/2022 3:12 PM, Richard Damon wrote:
>>>>>>>>>>>>> On 2/13/22 3:18 PM, olcott wrote:
>>>>>>>>>>>>>> On 2/13/2022 1:57 PM, Richard Damon wrote:
>>>>>>>>>>>>>>> On 2/13/22 2:43 PM, olcott wrote:
>>>>>>>>>>>>>>>> On 2/13/2022 11:19 AM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 2/13/22 9:38 AM, olcott wrote:
>>>>>>>>>>>>>>>>>> On 2/13/2022 5:43 AM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>> On 2/13/22 12:54 AM, olcott wrote:
>>>>>>>>>>>>>>>>>>>> On 2/12/2022 8:22 PM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>> On 2/12/22 9:02 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>> On 2/12/2022 7:54 PM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>> On 2/12/22 8:27 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/2022 6:57 PM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/22 7:37 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/2022 6:25 PM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/22 6:48 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/2022 5:39 PM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/22 5:34 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/2022 8:41 AM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/22 9:08 AM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/2022 6:49 AM, Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/12/22 12:01 AM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/2022 10:50 PM, Richard Damon
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/22 11:36 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/2022 6:58 PM, Richard Damon
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/22 7:52 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/2022 6:17 PM, Richard
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/22 7:10 PM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/2022 5:36 PM, Richard
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/22 10:20 AM, olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/11/2022 5:36 AM, Richard
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/10/22 11:39 PM, olcott
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/10/2022 10:20 PM,
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/10/22 10:58 PM, olcott
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/10/2022 6:02 PM,
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Richard Damon wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> On 2/10/22 9:18 AM,
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> olcott wrote:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>   > I explain how I am
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> necessarily correct on
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> the basis of the meaning
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> of my
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> words and you disagree
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> on the basis of your
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> failure to correctly
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> understand the meaning
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> of these words.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> No, you CLAIM to explain
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> based on the meaning of
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> the words, but use the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> wrong meaning of the words.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> THIS IS PROVEN TO BE
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> COMPLETELY TRUE ENTIRELY
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> ON THE BASIS OF THE
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> MEANING OF ITS WORDS:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> When a simulating halt
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> decider correctly
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> determines in a finite
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> number of steps that the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> pure UTM simulation of
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> its input would never
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> reach the final state of
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> this input then it can
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> correctly reject this
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> input as non-halting.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> IF it correctly decided,
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> then yes.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> But it has been shown, by
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> the definition of the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> construction method of H^
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> that if H <H^> <H^> goes
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> to H.Qn then H^ <H^> goes
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> to H^.Qn and Halts, and
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> thus by the definition of
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> a UTM, then we also have
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> that UTM <H^> <H^> will
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> halt.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Ĥ.q0 ⟨Ĥ⟩ ⊢* Ĥ.qx ⟨Ĥ⟩ ⟨Ĥ⟩
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> ⊢* Ĥ.qy ∞
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Ĥ.q0 ⟨Ĥ⟩ ⊢* Ĥ.qx ⟨Ĥ⟩ ⟨Ĥ⟩
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> ⊢* Ĥ.qn
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> You keep getting confused
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> between two things:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> (1) The execution of Ĥ ⟨Ĥ⟩
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> versus
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H ⟨Ĥ⟩ ⟨Ĥ⟩ (we
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> only look at the latter).
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> No, YOU qre confused.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> By the definioon of how to
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> build H^, embedded_H MUST
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> be EXACTLY the same
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> algorithm as H,
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> No it is specifically not
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> allowed to be.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H must have an
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> infinite loop appended to
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> its Ĥ.qy state and H is not
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> allowed to have such a loop
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> appended to its H.qy state.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Which is OUTSIDE the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> algorithm of H itself, and
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> doesn't affect the behavior
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> of H in deciding to go from
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Q0 to Qy/Qn.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> When the simulating halt
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> decider bases it halt status
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> decision on whether or not the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> same function is being called
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> with the same inputs the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> difference between H and
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H can change the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> behavior. A string comparison
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> between the machine
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> description of H and
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H yields false.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> No, it can't, not and be a
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> COMPUTATION.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> You obviously don't know the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> meaning of the words, so you
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> are just wrong.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Computation means for ALL
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> copoies, Same input leads to
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Same Output.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> The fact that it is not an exact
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> copy makes a difference.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> But it IS.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> H applied to ⟨Ĥ⟩ ⟨Ĥ⟩ does not
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> determine that itself is being
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> called multiple times with the
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> same input. embedded_H does
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> determine that itself is called
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> multiple times with the same input
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> because strcmp(H, embedded_H != 0.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Except that embedded_H does't have
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> a 'representation' of itself to use
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> to make that comparison, so that is
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> just more of your Fairy Dust
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Powered Unicorn stuff.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> It can very easily have a
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> representation of itself, that only
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> requires that it has access to its
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> own machine description.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> First, Turing Machines DON'T have
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> access to their own machine
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> description, unless it has been
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> provided as an input
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> You said that this was impossible.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> So, you are agreeing that they can't?
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Or do you just not understand the logic.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> I am agreeing that you contradicted
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> yourself.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> How, by saying that the only way a Turing
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Machine can have a copy of its
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> representation is for it to be given (and
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> H is defined in a way that it can't be
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> given as an extra input)?
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> No appended infinite loop making H and
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H the same
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Ḧ.q0 ⟨Ḧ⟩ ⊢* Ḧ.qx ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qy
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Ḧ.q0 ⟨Ḧ⟩ ⊢* Ḧ.qx ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> H ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* H.qn
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Except that the infinite loop isn't part of
>>>>>>>>>>>>>>>>>>>>>>>>>>>>> the copy of H in H^, it is something ADD to
>>>>>>>>>>>>>>>>>>>>>>>>>>>>> it, which only has/affects behavior AFTER H
>>>>>>>>>>>>>>>>>>>>>>>>>>>>> makes its decision.
>>>>>>>>>>>>>>>>>>>>>>>>>>>> So another words hypothetical examples get
>>>>>>>>>>>>>>>>>>>>>>>>>>>> you so confused you totally lose track of
>>>>>>>>>>>>>>>>>>>>>>>>>>>> everything.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>> What 'Hypothetical' are you referencing.
>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>> This is what I mean when I say that you hardly
>>>>>>>>>>>>>>>>>>>>>>>>>> pay any attention at all.
>>>>>>>>>>>>>>>>>>>>>>>>>> This is the hypothetical that I am referencing:
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>> No appended infinite loop making H and
>>>>>>>>>>>>>>>>>>>>>>>>>> embedded_H the same
>>>>>>>>>>>>>>>>>>>>>>>>>> Ḧ.q0 ⟨Ḧ⟩ ⊢* Ḧ.qx ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qy
>>>>>>>>>>>>>>>>>>>>>>>>>> Ḧ.q0 ⟨Ḧ⟩ ⊢* Ḧ.qx ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>>> And what do you mean by that?
>>>>>>>>>>>>>>>>>>>>>>>> When I redefine Ĥ to become Ḧ by eliminating its
>>>>>>>>>>>>>>>>>>>>>>>> infinite loop, I probably mean: {I redefine Ĥ to
>>>>>>>>>>>>>>>>>>>>>>>> become Ḧ by eliminating its infinite loop}.
>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> Which does what? Since if you aren't talking
>>>>>>>>>>>>>>>>>>>>>>> about Linz's H^, your results don't mean anything
>>>>>>>>>>>>>>>>>>>>>>> for the Halting Problem.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> It provides a bridge of understanding to my HP
>>>>>>>>>>>>>>>>>>>>>> refutation.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> The key skill that I have applied throughout my
>>>>>>>>>>>>>>>>>>>>>> career is eliminating "inessential complexity"
>>>>>>>>>>>>>>>>>>>>>> (1999 Turing award winner Fred Brooks) to make
>>>>>>>>>>>>>>>>>>>>>>   enormously difficult problems as simple as
>>>>>>>>>>>>>>>>>>>>>> possible.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> https://en.wikipedia.org/wiki/No_Silver_Bullet
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> But if you eliminate KEY ASPECTS then you aren't
>>>>>>>>>>>>>>>>>>>>> talking about what you need to.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> It is just like learning arithmetic before attacking
>>>>>>>>>>>>>>>>>>>> algebra.
>>>>>>>>>>>>>>>>>>>> It lays the foundation of prerequisites for my
>>>>>>>>>>>>>>>>>>>> actual rebuttal.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Just beware that if you make a statement that is only
>>>>>>>>>>>>>>>>>>> true for a limited case and don't explicitly state
>>>>>>>>>>>>>>>>>>> so, pointing that out is NOT a 'Dishonest Dodge'.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Since it is know that you are working on trying to
>>>>>>>>>>>>>>>>>>> prove that a Unicorn exists, what you are saying WILL
>>>>>>>>>>>>>>>>>>> be looked at in a light anticipating where you are
>>>>>>>>>>>>>>>>>>> going.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Also remember that showing a rule happens to be
>>>>>>>>>>>>>>>>>>> correct for one case, doesn't prove that it will be
>>>>>>>>>>>>>>>>>>> for a different case.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> You have gone through all of this before, and it came
>>>>>>>>>>>>>>>>>>> for nothing, but if this is how you want to spend
>>>>>>>>>>>>>>>>>>> your last days, knock yourself out.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> I think that you already understand that
>>>>>>>>>>>>>>>>>> With Ĥ ⟨Ĥ⟩
>>>>>>>>>>>>>>>>>> Ĥ.q0 ⟨Ĥ⟩ ⊢* Ĥ.qx ⟨Ĥ⟩ ⟨Ĥ⟩ ⊢* Ĥ.qy ∞ // this path is
>>>>>>>>>>>>>>>>>> never taken
>>>>>>>>>>>>>>>>>> Ĥ.q0 ⟨Ĥ⟩ ⊢* Ĥ.qx ⟨Ĥ⟩ ⟨Ĥ⟩ ⊢* Ĥ.qn
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> making Ḧ ⟨Ḧ⟩ an equivalent computation
>>>>>>>>>>>>>>>>>> Ḧ.q0 ⟨Ḧ⟩ ⊢* Ḧ.qx ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qy
>>>>>>>>>>>>>>>>>> Ḧ.q0 ⟨Ḧ⟩ ⊢* Ḧ.qx ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> And, as seems common with your arguments, you keep on
>>>>>>>>>>>>>>>>> forgetting the CONDITIONS on each line.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> H^ <H^> is
>>>>>>>>>>>>>>>>> H^.q0 <H^> -> H^.Qx <H^> <H^>
>>>>>>>>>>>>>>>>> Then
>>>>>>>>>>>>>>>>> H^.Qx <H^> <H^> -> H^.Qy -> ∞ IF and only if H <H^>
>>>>>>>>>>>>>>>>> <H^> -> H.Qy and
>>>>>>>>>>>>>>>>> H^.Qx <H^> <H^> -> H^.Qn If and ohly if H <H^> <H^> =>
>>>>>>>>>>>>>>>>> H.Qn.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> If you stipulate that H <H^> <H^> will never go to
>>>>>>>>>>>>>>>>> H.Qy, then the behavior on that path can be changed
>>>>>>>>>>>>>>>>> with no effect.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Without the If and only if clauses, the initial
>>>>>>>>>>>>>>>>> description is incorrect because it is incomplete.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> So, WITH THE STIPULATION THAT H won't go to H.Qy for
>>>>>>>>>>>>>>>>> either version, then changing H^ to the H" that omits
>>>>>>>>>>>>>>>>> to loop is an equivalence, but ONLY under that
>>>>>>>>>>>>>>>>> stipulation.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> This of course shows that H will be wrong about H", as
>>>>>>>>>>>>>>>>> H" will ALWAYS Halt if H answers, and H not answering
>>>>>>>>>>>>>>>>> is always wrong. Thus H will either be wrong for not
>>>>>>>>>>>>>>>>> answering or giving the wrong answer.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I am only making two versions of input to H:
>>>>>>>>>>>>>>>> (1) Ĥ WITH an appended infinite loop
>>>>>>>>>>>>>>>> (2) Ḧ WITHOUT an appended infinite loop
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Only this is being examined:
>>>>>>>>>>>>>>>> embedded_H ⟨Ĥ⟩ ⟨Ĥ⟩
>>>>>>>>>>>>>>>> embedded_H ⟨Ḧ⟩ ⟨Ḧ⟩
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Right, but the conclusion that H" is 'equivalent' to H^
>>>>>>>>>>>>>>> is only true (if you mean equivalent in the sense that
>>>>>>>>>>>>>>> they compute the same function) if it is the case that
>>>>>>>>>>>>>>> neither of H <H^> <H^> or H <H"> <H"> go to H.Qy.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Good.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Unless you want to indicate some other definition of
>>>>>>>>>>>>>>> 'equivalent' you using (that could be proper to do so
>>>>>>>>>>>>>>> here), you need to include the conditions under which the
>>>>>>>>>>>>>>> statement is true.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> embedded_H ⟨Ĥ⟩ ⟨Ĥ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>    and
>>>>>>>>>>>>>> embedded_H ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>    nd
>>>>>>>>>>>>>> H ⟨Ḧ⟩ ⟨Ḧ⟩ ⊢* Ḧ.qn
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> Problem, embedded_H / H need to transition to a state in H,
>>>>>>>>>>>>> not some other machine.
>>>>>>>>>>>>
>>>>>>>>>>>> As soon as we append an infinite loop to H.y is it no longer H.
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> This is where you are showing your lack of understanding of
>>>>>>>>>>> Turing Machines.
>>>>>>>>>>>
>>>>>>>>>>> NO ONE has said that the machine where we added the loop is
>>>>>>>>>>> still the machine H, in fact, Linz calls that machine H', but
>>>>>>>>>>> H' CONTAINS a complete copy of H, and that copy will still
>>>>>>>>>>> act exactly like the original H to the point where it gets to
>>>>>>>>>>> the stat Qy.
>>>>>>>>>>>
>>>>>>>>>>> This ability to compose machines of copies of other machines
>>>>>>>>>>> is basically like the concept of calling subroutines (even if
>>>>>>>>>>> it is implemented differently) and is fundamental to the
>>>>>>>>>>> design and analysis of Turing Macines.
>>>>>>>>>>>
>>>>>>>>>>> Would you have a problem saying the subroutine H is no longer
>>>>>>>>>>> the subroutine H if one function just calls H and returns
>>>>>>>>>>> while a second calls H and conditionally loops? Yes, the
>>>>>>>>>>> whole program is not H, but the subroutine H is still there
>>>>>>>>>>> and will behave exactly like it used to in both of the cases.
>>>>>>>>>>>
>>>>>>>>>>> One way to map a Turing Machine to ordinary software is to
>>>>>>>>>>> think of the Q0 state (or whatever is the 'starting' state of
>>>>>>>>>>> the Turing machine) as the entry point for the function, and
>>>>>>>>>>> the Halting States of the Turing Machine as retrun stateents
>>>>>>>>>>> which return a value indicating what state the machine ended in.
>>>>>>>>>>>
>>>>>>>>>>> Thus the modifications Linz has done to H are nothing more
>>>>>>>>>>> that building H^ as mostly a call to H, with code before the
>>>>>>>>>>> call to manipulate the tape to add the second copy, and code
>>>>>>>>>>> after the return to loop forever if H returns the 'Halting'
>>>>>>>>>>> answer.
>>>>>>>>>>>
>>>>>>>>>>> The Machine/Subroutine H has not been touched at all.
>>>>>>>>>>
>>>>>>>>>> My point is that Ĥ ⟨Ĥ⟩ is equivalent to Ḧ ⟨Ḧ⟩
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> And my point is that they are only equivalent in the normal
>>>>>>>>> sense of the word if neither of H <H^> <H^> and H <H"> <H"> go
>>>>>>>>> to H.Qy
>>>>>>>>>
>>>>>>>>> Without that qualification, it is a false statement. PERIOD.
>>>>>>>>
>>>>>>>> They are equivalent in that neither can possibly go to their q.y
>>>>>>>> state.
>>>>>>>
>>>>>>>
>>>>>>> THAT is incorrect without the same
>>>>>>> qualification/assumption/stipulation, the H doesn't go to Qy.
>>>>>>>
>>>>>>
>>>>>> If H was a white cat detector and you presented H with a black cat
>>>>>> would it say "yes" ?
>>>>>>
>>>>>
>>>>> But we aren't talking about a 'detector'
>>>>
>>>> Sure we are.
>>>>
>>>>
>>>
>>> Then you don't know what you are talking about (and showing your
>>> dishonesty by your clipping).
>>>
>>> THe claim that H^ and H" are equivilent machines has NOTHING to do
>>> with there being a 'Detector' but do they behave exactly the same.
>> So in other words you are disavowing that both ⟨Ĥ⟩ and ⟨Ḧ⟩ have a copy
>> of H embedded within them ?
>>
>> A halt detector is the same idea as a halt decider yet a halt detector
>> need not get every input correctly. Every input that the halt detector
>> gets correctly is in the domain of the computable function that it
>> implements.
>>
>>
>
> Yes, they have a copy of the Halt Detector H in them, and unless you are
> willing to stipulate that H will not go to H.Qy when given H^ or H" as
> an input, then you can not show that those machines are equivalent.

If the simulating halt decider H cannot possibly go to H.qy on a
specific input then any such stipulation would be redundant for this input.

--
Copyright 2021 Pete Olcott

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

SubjectRepliesAuthor
o Concise refutation of halting problem proofs V62 [ Linz Proof ]

By: olcott on Sun, 6 Feb 2022

163olcott
server_pubkey.txt

rocksolid light 0.9.8
clearnet tor