Warning: mysqli::__construct(): (HY000/1203): User howardkn already has more than 'max_user_connections' active connections in D:\Inetpub\vhosts\howardknight.net\al.howardknight.net\includes\artfuncs.php on line 21
Failed to connect to MySQL: (1203) User howardkn already has more than 'max_user_connections' active connectionsPath: news.eternal-september.org!eternal-september.org!feeder3.eternal-september.org!news.quux.org!news.nk.ca!rocksolid2!i2pn2.org!.POSTED!not-for-mail
From: Richard Damon
Newsgroups: comp.theory
Subject: Re: How computable functions actually work. (was Flibble)
Date: Tue, 22 Apr 2025 22:33:11 -0400
Organization: i2pn2 (i2pn.org)
Message-ID:
References:
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Injection-Date: Wed, 23 Apr 2025 02:34:22 -0000 (UTC)
Injection-Info: i2pn2.org;
logging-data="1482570"; mail-complaints-to="usenet@i2pn2.org";
posting-account="diqKR1lalukngNWEqoq9/uFtbkm5U+w3w6FQ0yesrXg";
User-Agent: Mozilla Thunderbird
In-Reply-To:
X-Spam-Checker-Version: SpamAssassin 4.0.0
Content-Language: en-US
On 4/22/25 6:19 PM, olcott wrote:
> On 4/22/2025 4:58 PM, Andy Walker wrote:
>> On 22/04/2025 15:57, Mr Flibble wrote:
>>> On Tue, 22 Apr 2025 15:43:27 +0100, Andy Walker wrote:
>>>> The "real" Mr Flibble is a malevolent penguin. I wonder why
>>>> contributors take him so seriously? If you want to debate with a
>>>> penguin, that's your prerogative, but to me it makes more sense to add
>>>> several pinches of salt and smile or groan as appropriate to everything
>>>> he writes. He has a knack for writing things that are just about
>>>> plausible, which is enviable, but one response to anything interesting
>>>> is surely enough?
>>> Mr Flibble is very cross.
>>
>> He shouldn't be. As hinted above, being able to write successful
>> satire is a rare skill. But it loses its point if too many people take
>> it seriously.
>>
>
> Flibble factually correct.
>
> All computation is defined to be represented as finite string
> transformations to finite strings.
Except you are doing the logic backward.
>
> This how Turing Machine computable functions actually work.
> Outputs are forced to correspond to inputs when finite string
> transformation rules are applied to inputs to derive outputs.
And you need to know that the function *IS* computable to use that.
What the machine actually produces will be computable.
What the machine is SUPPOSED to produce might not be.
>
> a function is computable if there exists an algorithm
> that can do the job of the function, i.e. given an input
> of the function domain it can return the corresponding
> output. https://en.wikipedia.org/wiki/Computable_function
Which says if a machine exists, it is conputable.
The machine does not need to exist.
This seems to be a flaw in your logic, you seem to think there is a
Truth Faerie that can magically make the impossible happen.
>
> On Turing Machines inputs finite strings, and
> finite string transformation rules applied to
> these finite strings to derive corresponding outputs.
Yes, so the results can only BE what is computable, but as pointed out,
the correct answer need not be.
>
> People here stupidly assume that the outputs are not
> required to correspond to the inputs. That comes from
> learn-by-rote with zero depth of understanding.
>
The outputs DO need to correspond to the input, but not necessarily by a
computable transform.
That only exists if the function is, in fact, computable.
Nearly all "random" functions will not be, by a simple counting
argument, There are more functions/mapping (Aleph-1) that can be asked
for than machines (Aleph-0) to do the mapping. Therefore only a
infintisimal fraction of all possible mappings are in fact computable.
Now, when we limit ourselves to mapping that have a use, it seems we
find a lot that do have a mapping, but not all.
There is no computation that can compute for *ALL* possible programs, if
that progran will halt when run. Turing proves that by showing for every
possible decider that we might think can do it, that we can make a
program for it to decide on that it will get wrong.
Thus the mapping must be uncomputable.