On 2/19/24 10:32 PM, olcott wrote:
> On 2/19/2024 9:28 PM, immibis wrote:
>> On 20/02/24 04:19, olcott wrote:
>>> On 2/19/2024 8:13 PM, immibis wrote:
>>>> On 20/02/24 02:20, olcott wrote:
>>>>> On 2/19/2024 7:14 PM, immibis wrote:
>>>>>> On 20/02/24 01:19, olcott wrote:
>>>>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> What would you expect someone who was paying attention
>>>>>>>>>>>>>>>>>> to see?
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions
>>>>>>>>>>>>>>>>> that must
>>>>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>>>>
>>>>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>
>>>>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>>>>
>>>>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>>>>> stronger
>>>>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>>>>> its own
>>>>>>>>>>> description.
>>>>>>>>>>
>>>>>>>>>> OBviosly you haven't been reading.
>>>>>>>>>>
>>>>>>>>>> The Human mind is not limited by the deterministic
>>>>>>>>>> requirement, or the requirement to only use the designated input.
>>>>>>>>>>
>>>>>>>>>> Humans have "Free-Will", while Turing Macines are
>>>>>>>>>> "Deterministic".
>>>>>>>>>>
>>>>>>>>>> Totally different.
>>>>>>>>>>
>>>>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>>>>> enslaved on a predetermined path that you are stuck on forever
>>>>>>>>>> now.
>>>>>>>>>
>>>>>>>>> Show me a technical paper that attempts to prove that the
>>>>>>>>> human mind can perform computations that will remains forever
>>>>>>>>> impossible for any advance in technological innovation of
>>>>>>>>> computations.
>>>>>>>>>
>>>>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>>>>
>>>>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>>>>> random number generator. In fact, a random number generator is
>>>>>>>> an essential part of the algorithm that generates text using a
>>>>>>>> GPT model. If you change the random number generator with a
>>>>>>>> non-random number generator, that always returns 0.4 for
>>>>>>>> example, then it will be deterministic.
>>>>>>>
>>>>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>>>>
>>>>>>
>>>>>> LLMs do not have minds. When you understand how they work, it is
>>>>>> obvious that they are deterministic unless a random number
>>>>>> generator is used.
>>>>>
>>>>> Chat-GPT4 was able to get a progressively deeper understanding
>>>>> of terminology and the subject matter as the conversation proceeded.
>>>>> It could do this only because it is dynamically self-programmed.
>>>>>
>>>> You do not understand how GPT models work. When you understand how
>>>> they work, you will know how stupid this statement is.
>>>>
>>>> Speaking extremely loosely (if you want a precise statement, then
>>>> LEARN HOW THEY WORK) GPT can search the conversation so far to find
>>>> relevant facts.
>>>
>>> ChatGPT-4 did continually learn numerous new things within
>>> a single dialogue.
>>>
>> Yes, it's able to search the conversation for more than one fact.
>
> It incrementally built an increasingly accurate model of
> the subject matter over and above its training.
>

On 2/19/24 11:02 AM, olcott wrote:
> On 2/19/2024 1:06 AM, immibis wrote:
>> On 19/02/24 07:42, olcott wrote:
>>> On 2/19/2024 12:29 AM, immibis wrote:
>>>> On 19/02/24 07:11, olcott wrote:
>>>>> On 2/18/2024 8:40 PM, immibis wrote:
>>>>>> On 19/02/24 02:28, olcott wrote:
>>>>>>> On 2/18/2024 6:27 PM, Richard Damon wrote:
>>>>>>>> On 2/18/24 6:47 PM, olcott wrote:
>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>
>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor infinite.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>> no
>>>>>>>>>>>>>
>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>
>>>>>>>>>>>> What would you expect someone who was paying attention to see?
>>>>>>>>>>>
>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>
>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>
>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>
>>>>>>>>> "This sentence is not true."
>>>>>>>>>
>>>>>>>>> *Increasing specificity*
>>>>>>>>> (1) Says something about something.
>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>> (3) Says that something is not true.
>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>> But is isn't the question being asked, as the compuation that is
>>>>>>>> deciding is H, and the input is not just a description of H.
>>>>>>>>
>>>>>>>
>>>>>>> Generically:
>>>>>>> "Does the Computation described by this input Halt?"
>>>>>>
>>>>>> That's right. Does it halt?
>>>>>>
>>>>>>>
>>>>>>> Specifically Ȟ is being asked:
>>>>>>> Do you halt on your own Turing Machine description?
>>>>>>
>>>>>> No, it's a different description.
>>>>>
>>>>> // *Original Linz H with simpler syntax*
>>>>> H.q0 ⟨H⟩ ⟨H⟩ ⊢* H.qy  // H applied to ⟨H⟩ halts
>>>>> H.q0 ⟨H⟩ ⟨H⟩ ⊢* H.qn  // H applied to ⟨H⟩ does not halt
>>>>> H correctly transitions to H.qy
>>>>>
>>>>> // *The self-contradictory version of H*
>>>>> Ȟ.q0 ⟨Ȟ⟩ ⟨Ȟ⟩ ⊢* Ȟ.qy ∞ // Ȟ applied to ⟨Ȟ⟩ halts
>>>>> Ȟ.q0 ⟨Ȟ⟩ ⟨Ȟ⟩ ⊢* Ȟ.qn   // Ȟ applied to ⟨Ȟ⟩ does not halt
>>>>> Neither Yes nor No is the correct answer.
>>>>>
>>>>>
>>>>>
>>>>
>>>> It is trivial to change any machine description to a different
>>>> description of a different machine which always returns the same
>>>> result as the original machine for any input.
>>>
>>> *H and its self-contradictory*
>>> *version Ȟ prove that I am correct*
>>>
>> H and its self-contradictory version Ȟ do not prove that some
>> sequences aren't finite or infinite.
>
> When I make a point dishonest people change the subject.

No, you make stupid statments that people respond to, and when you don't
like what they say, you claim they are changing the subject, and thus LIE.

> *It is true that Ȟ <is> the self-contradictory version of H*
> The price of tea in China will not change this.
>

So, you think that altering something keeps it the same?

You are just proving that you are either totally stupid to think
something like that is true, or a pathological liar that can't see what
truth actually is, or both. (or perhaps you are just insai)

Ȟ doesn't meet the basic requirements to BE an "H", so it can't be a
"version" of it.

I guess you are admitting that you are just a "version" of Donald Trump.

You use the same sort of logic as him, so I guess that makes you just
versions of each other.

On 2/19/24 7:21 PM, olcott wrote:
> On 2/19/2024 6:11 PM, immibis wrote:
>> On 19/02/24 17:02, olcott wrote:
>>> On 2/19/2024 1:06 AM, immibis wrote:
>>>> H and its self-contradictory version Ȟ do not prove that some
>>>> sequences aren't finite or infinite.
>>>
>>> When I make a point dishonest people change the subject.
>>> *It is true that Ȟ <is> the self-contradictory version of H*
>>> The price of tea in China will not change this.
>>>
>>
>> The subject is whether D(D) has a finite or infinite execution sequence.
>>
>
> The subject is why can't a halt decider exist?
> Because semantically unsound inputs are not allowed to be rejected.
>
> What can't a Truth predicate exist?
> Because semantically unsound inputs are not allowed to be rejected.
>

So, you AGREE that Halt Deciders can't exist?

Then why do you claim otherwise?

Since the actual Halting question is valid, all you have done is proved
that your POOP is invalid.

You just don't know what "Semantically Valid" means, because you seem to
have enormous misconceptions about so many things.

On 2/19/24 8:21 PM, olcott wrote:
> On 2/19/2024 7:15 PM, immibis wrote:
>> On 20/02/24 01:21, olcott wrote:
>>> On 2/19/2024 6:11 PM, immibis wrote:
>>>> On 19/02/24 17:02, olcott wrote:
>>>>> On 2/19/2024 1:06 AM, immibis wrote:
>>>>>> H and its self-contradictory version Ȟ do not prove that some
>>>>>> sequences aren't finite or infinite.
>>>>>
>>>>> When I make a point dishonest people change the subject.
>>>>> *It is true that Ȟ <is> the self-contradictory version of H*
>>>>> The price of tea in China will not change this.
>>>>>
>>>>
>>>> The subject is whether D(D) has a finite or infinite execution
>>>> sequence.
>>>>
>>>
>>> The subject is why can't a halt decider exist?
>>
>> If a halt decider exists, then D(D) has a finite or infinite execution
>> sequence.
>
> Continuing with a dishonest change-the-subject
> rebuttal because you are only a dishonest troll.
>
>

Good description of your whole argument.

PO just acting like DT.

On 2/19/24 8:27 PM, olcott wrote:
> On 2/19/2024 7:15 PM, immibis wrote:
>> On 20/02/24 01:21, olcott wrote:
>>> On 2/19/2024 6:11 PM, immibis wrote:
>>>> On 19/02/24 17:02, olcott wrote:
>>>>> On 2/19/2024 1:06 AM, immibis wrote:
>>>>>> H and its self-contradictory version Ȟ do not prove that some
>>>>>> sequences aren't finite or infinite.
>>>>>
>>>>> When I make a point dishonest people change the subject.
>>>>> *It is true that Ȟ <is> the self-contradictory version of H*
>>>>> The price of tea in China will not change this.
>>>>>
>>>>
>>>> The subject is whether D(D) has a finite or infinite execution
>>>> sequence.
>>>>
>>>
>>> The subject is why can't a halt decider exist?
>>
>> If a halt decider exists, then D(D) has a finite or infinite execution
>> sequence.
>
> We can equally determine that no baker exists because no baker
> can bake an angel food cakes using only house bricks for ingredients.
> When we ask: What can no baker exists? The answer is that the problem
> definition is incorrect.
>

Another of your "Dishonest Dodges"

You do seem to like you herring with red sauce.

I guess your problem is you don't know enough of how Turing Machines or
Computations work to make an intelegent arguement, so you need to use
dishonest dodges when you get cornered.

You are just proving you utter ignorance and pathological lying.

On 20/02/24 04:47, Richard Damon wrote:
> On 2/19/24 7:19 PM, olcott wrote:
>> On 2/19/2024 6:10 PM, immibis wrote:
>>> On 19/02/24 16:46, olcott wrote:
>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor infinite.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>
>>>>>>>>>>>>> What would you expect someone who was paying attention to see?
>>>>>>>>>>>>
>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>
>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>
>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>
>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>
>>>>>>>>>> *Increasing specificity*
>>>>>>>>>> (1) Says something about something.
>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>
>>>>>>>>> "Does this computation halt?"
>>>>>>>>>
>>>>>>>>> are precise questions which are not vague.
>>>>>>>>
>>>>>>>> A human can see that this is the question:
>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>
>>>>>>> A Turing machine has no way to access its own description.
>>>>>>
>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>> stronger
>>>>>> than a TM. Therefore a TM can determine that it is processing its own
>>>>>> description.
>>>>>
>>>>> OBviosly you haven't been reading.
>>>>>
>>>>> The Human mind is not limited by the deterministic requirement, or
>>>>> the requirement to only use the designated input.
>>>>>
>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>
>>>>> Totally different.
>>>>>
>>>>> Maybe YOU have lost you ability to independently think, having
>>>>> traded your free-will to the devil for something, and are now
>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>
>>>> Show me a technical paper that attempts to prove that the
>>>> human mind can perform computations that will remains forever
>>>> impossible for any advance in technological innovation of
>>>> computations.
>>>>
>>>> LLMs have already broken through any limitation of determinism.
>>>
>>> LLMs are deterministic, actually, unless you connect them to a random
>>> number generator. In fact, a random number generator is an essential
>>> part of the algorithm that generates text using a GPT model. If you
>>> change the random number generator with a non-random number
>>> generator, that always returns 0.4 for example, then it will be
>>> deterministic.
>>
>> LLMs can change their mind in the middle of a single dialogue.
>>
>
> They might respond differently then they did before, but that is because
> each "input" isn't a "fresh" computation, but a continuation of the
> previous one.
>
> "Conversations" don't really fit into the computation model.

Actually, each input is a fresh computation, with the previous
conversation history as input.

Actually, each token is a fresh computation, with the previous
conversation history as input. A token is usually about a syllable long.

On 20/02/24 04:32, olcott wrote:
> On 2/19/2024 9:28 PM, immibis wrote:
>> On 20/02/24 04:19, olcott wrote:
>>> On 2/19/2024 8:13 PM, immibis wrote:
>>>> On 20/02/24 02:20, olcott wrote:
>>>>> On 2/19/2024 7:14 PM, immibis wrote:
>>>>>> On 20/02/24 01:19, olcott wrote:
>>>>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> What would you expect someone who was paying attention
>>>>>>>>>>>>>>>>>> to see?
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions
>>>>>>>>>>>>>>>>> that must
>>>>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>>>>
>>>>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>
>>>>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>>>>
>>>>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>>>>> stronger
>>>>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>>>>> its own
>>>>>>>>>>> description.
>>>>>>>>>>
>>>>>>>>>> OBviosly you haven't been reading.
>>>>>>>>>>
>>>>>>>>>> The Human mind is not limited by the deterministic
>>>>>>>>>> requirement, or the requirement to only use the designated input.
>>>>>>>>>>
>>>>>>>>>> Humans have "Free-Will", while Turing Macines are
>>>>>>>>>> "Deterministic".
>>>>>>>>>>
>>>>>>>>>> Totally different.
>>>>>>>>>>
>>>>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>>>>> enslaved on a predetermined path that you are stuck on forever
>>>>>>>>>> now.
>>>>>>>>>
>>>>>>>>> Show me a technical paper that attempts to prove that the
>>>>>>>>> human mind can perform computations that will remains forever
>>>>>>>>> impossible for any advance in technological innovation of
>>>>>>>>> computations.
>>>>>>>>>
>>>>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>>>>
>>>>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>>>>> random number generator. In fact, a random number generator is
>>>>>>>> an essential part of the algorithm that generates text using a
>>>>>>>> GPT model. If you change the random number generator with a
>>>>>>>> non-random number generator, that always returns 0.4 for
>>>>>>>> example, then it will be deterministic.
>>>>>>>
>>>>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>>>>
>>>>>>
>>>>>> LLMs do not have minds. When you understand how they work, it is
>>>>>> obvious that they are deterministic unless a random number
>>>>>> generator is used.
>>>>>
>>>>> Chat-GPT4 was able to get a progressively deeper understanding
>>>>> of terminology and the subject matter as the conversation proceeded.
>>>>> It could do this only because it is dynamically self-programmed.
>>>>>
>>>> You do not understand how GPT models work. When you understand how
>>>> they work, you will know how stupid this statement is.
>>>>
>>>> Speaking extremely loosely (if you want a precise statement, then
>>>> LEARN HOW THEY WORK) GPT can search the conversation so far to find
>>>> relevant facts.
>>>
>>> ChatGPT-4 did continually learn numerous new things within
>>> a single dialogue.
>>>
>> Yes, it's able to search the conversation for more than one fact.
>
> It incrementally built an increasingly accurate model of
> the subject matter over and above its training.
>

On 2/19/2024 10:00 PM, immibis wrote:
> On 20/02/24 04:47, Richard Damon wrote:
>> On 2/19/24 7:19 PM, olcott wrote:
>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>> On 19/02/24 16:46, olcott wrote:
>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor infinite.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> What would you expect someone who was paying attention to
>>>>>>>>>>>>>> see?
>>>>>>>>>>>>>
>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>
>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>
>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>
>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>
>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>
>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>
>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>
>>>>>>>>> A human can see that this is the question:
>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>
>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>
>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>> stronger
>>>>>>> than a TM. Therefore a TM can determine that it is processing its
>>>>>>> own
>>>>>>> description.
>>>>>>
>>>>>> OBviosly you haven't been reading.
>>>>>>
>>>>>> The Human mind is not limited by the deterministic requirement, or
>>>>>> the requirement to only use the designated input.
>>>>>>
>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>
>>>>>> Totally different.
>>>>>>
>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>> traded your free-will to the devil for something, and are now
>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>
>>>>> Show me a technical paper that attempts to prove that the
>>>>> human mind can perform computations that will remains forever
>>>>> impossible for any advance in technological innovation of
>>>>> computations.
>>>>>
>>>>> LLMs have already broken through any limitation of determinism.
>>>>
>>>> LLMs are deterministic, actually, unless you connect them to a
>>>> random number generator. In fact, a random number generator is an
>>>> essential part of the algorithm that generates text using a GPT
>>>> model. If you change the random number generator with a non-random
>>>> number generator, that always returns 0.4 for example, then it will
>>>> be deterministic.
>>>
>>> LLMs can change their mind in the middle of a single dialogue.
>>>
>>
>> They might respond differently then they did before, but that is
>> because each "input" isn't a "fresh" computation, but a continuation
>> of the previous one.
>>
>> "Conversations" don't really fit into the computation model.
>
> Actually, each input is a fresh computation, with the previous
> conversation history as input.
>

That seems correct. None-the-less it does get an increasingly
deeper understanding of the subject matter, thus is not at
all "deterministic" in any conventional use of the term.

> Actually, each token is a fresh computation, with the previous
> conversation history as input. A token is usually about a syllable long.

--
Copyright 2024 Olcott "Talent hits a target no one else can hit; Genius
hits a target no one else can see." Arthur Schopenhauer

On 2/19/24 11:00 PM, immibis wrote:
> On 20/02/24 04:47, Richard Damon wrote:
>> On 2/19/24 7:19 PM, olcott wrote:
>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>> On 19/02/24 16:46, olcott wrote:
>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor infinite.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> What would you expect someone who was paying attention to
>>>>>>>>>>>>>> see?
>>>>>>>>>>>>>
>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>
>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>
>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>
>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>
>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>
>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>
>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>
>>>>>>>>> A human can see that this is the question:
>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>
>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>
>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>> stronger
>>>>>>> than a TM. Therefore a TM can determine that it is processing its
>>>>>>> own
>>>>>>> description.
>>>>>>
>>>>>> OBviosly you haven't been reading.
>>>>>>
>>>>>> The Human mind is not limited by the deterministic requirement, or
>>>>>> the requirement to only use the designated input.
>>>>>>
>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>
>>>>>> Totally different.
>>>>>>
>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>> traded your free-will to the devil for something, and are now
>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>
>>>>> Show me a technical paper that attempts to prove that the
>>>>> human mind can perform computations that will remains forever
>>>>> impossible for any advance in technological innovation of
>>>>> computations.
>>>>>
>>>>> LLMs have already broken through any limitation of determinism.
>>>>
>>>> LLMs are deterministic, actually, unless you connect them to a
>>>> random number generator. In fact, a random number generator is an
>>>> essential part of the algorithm that generates text using a GPT
>>>> model. If you change the random number generator with a non-random
>>>> number generator, that always returns 0.4 for example, then it will
>>>> be deterministic.
>>>
>>> LLMs can change their mind in the middle of a single dialogue.
>>>
>>
>> They might respond differently then they did before, but that is
>> because each "input" isn't a "fresh" computation, but a continuation
>> of the previous one.
>>
>> "Conversations" don't really fit into the computation model.
>
> Actually, each input is a fresh computation, with the previous
> conversation history as input.

That is ONE way to look at it.

>
> Actually, each token is a fresh computation, with the previous
> conversation history as input. A token is usually about a syllable long.

No, most of the models take in the whole prompt before generating any of
the final output. Yes, SOME of the processing starts token by token, but
later parts can affect the begining of the output, so it does need to
digest the full input before the "computation" completes.

On 20/02/24 05:10, Richard Damon wrote:
> On 2/19/24 11:00 PM, immibis wrote:
>> On 20/02/24 04:47, Richard Damon wrote:
>>> On 2/19/24 7:19 PM, olcott wrote:
>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> What would you expect someone who was paying attention to
>>>>>>>>>>>>>>> see?
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>
>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>
>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>
>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>
>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>
>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>
>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>
>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>
>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>> stronger
>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>> its own
>>>>>>>> description.
>>>>>>>
>>>>>>> OBviosly you haven't been reading.
>>>>>>>
>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>> or the requirement to only use the designated input.
>>>>>>>
>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>
>>>>>>> Totally different.
>>>>>>>
>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>>
>>>>>> Show me a technical paper that attempts to prove that the
>>>>>> human mind can perform computations that will remains forever
>>>>>> impossible for any advance in technological innovation of
>>>>>> computations.
>>>>>>
>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>
>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>> random number generator. In fact, a random number generator is an
>>>>> essential part of the algorithm that generates text using a GPT
>>>>> model. If you change the random number generator with a non-random
>>>>> number generator, that always returns 0.4 for example, then it will
>>>>> be deterministic.
>>>>
>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>
>>>
>>> They might respond differently then they did before, but that is
>>> because each "input" isn't a "fresh" computation, but a continuation
>>> of the previous one.
>>>
>>> "Conversations" don't really fit into the computation model.
>>
>> Actually, each input is a fresh computation, with the previous
>> conversation history as input.
>
> That is ONE way to look at it.
>
>>
>> Actually, each token is a fresh computation, with the previous
>> conversation history as input. A token is usually about a syllable long.
>
> No, most of the models take in the whole prompt before generating any of
> the final output.

On 20/02/24 05:09, olcott wrote:
> On 2/19/2024 10:00 PM, immibis wrote:
>> On 20/02/24 04:47, Richard Damon wrote:
>>> On 2/19/24 7:19 PM, olcott wrote:
>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> What would you expect someone who was paying attention to
>>>>>>>>>>>>>>> see?
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>
>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>
>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>
>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>
>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>
>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>
>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>
>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>
>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>> stronger
>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>> its own
>>>>>>>> description.
>>>>>>>
>>>>>>> OBviosly you haven't been reading.
>>>>>>>
>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>> or the requirement to only use the designated input.
>>>>>>>
>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>
>>>>>>> Totally different.
>>>>>>>
>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>>
>>>>>> Show me a technical paper that attempts to prove that the
>>>>>> human mind can perform computations that will remains forever
>>>>>> impossible for any advance in technological innovation of
>>>>>> computations.
>>>>>>
>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>
>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>> random number generator. In fact, a random number generator is an
>>>>> essential part of the algorithm that generates text using a GPT
>>>>> model. If you change the random number generator with a non-random
>>>>> number generator, that always returns 0.4 for example, then it will
>>>>> be deterministic.
>>>>
>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>
>>>
>>> They might respond differently then they did before, but that is
>>> because each "input" isn't a "fresh" computation, but a continuation
>>> of the previous one.
>>>
>>> "Conversations" don't really fit into the computation model.
>>
>> Actually, each input is a fresh computation, with the previous
>> conversation history as input.
>>
>
> That seems correct. None-the-less it does get an increasingly
> deeper understanding of the subject matter, thus is not at
> all "deterministic" in any conventional use of the term.

immibis <news@immibis.com> writes:

> The subject is whether D(D) has a finite or infinite execution
> sequence.

I case it helps you decide the value of the interactions you are having
(I know it can be fun) PO has stated categorically that the wrong answer
is the right one:

Me: Here's the key question: do you still assert that H(P,P) == false is
the "correct" answer even though P(P) halts?

PO: Yes that is the correct answer even though P(P) halts.

And, so you know what you might be getting into he is prepared to say
things like this:

"the fact that a computation halts does not entail that it is a
halting computation"

and, when pressed, he is even prepared to state categorical opposites
and not correct either:

"Furthermore I have repeated H.q0 ⟨Ĥ⟩ ⟨Ĥ⟩ transitions to H.qn many
times."

"No nitwit H ⟨Ĥ⟩ ⟨Ĥ⟩ transitions to H.qy as I have told you many
times."

--
Ben.

On 2/19/24 11:09 PM, olcott wrote:
> On 2/19/2024 10:00 PM, immibis wrote:
>> On 20/02/24 04:47, Richard Damon wrote:
>>> On 2/19/24 7:19 PM, olcott wrote:
>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> What would you expect someone who was paying attention to
>>>>>>>>>>>>>>> see?
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>
>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>
>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>
>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>
>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>
>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>
>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>
>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>
>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>> stronger
>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>> its own
>>>>>>>> description.
>>>>>>>
>>>>>>> OBviosly you haven't been reading.
>>>>>>>
>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>> or the requirement to only use the designated input.
>>>>>>>
>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>
>>>>>>> Totally different.
>>>>>>>
>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>>
>>>>>> Show me a technical paper that attempts to prove that the
>>>>>> human mind can perform computations that will remains forever
>>>>>> impossible for any advance in technological innovation of
>>>>>> computations.
>>>>>>
>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>
>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>> random number generator. In fact, a random number generator is an
>>>>> essential part of the algorithm that generates text using a GPT
>>>>> model. If you change the random number generator with a non-random
>>>>> number generator, that always returns 0.4 for example, then it will
>>>>> be deterministic.
>>>>
>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>
>>>
>>> They might respond differently then they did before, but that is
>>> because each "input" isn't a "fresh" computation, but a continuation
>>> of the previous one.
>>>
>>> "Conversations" don't really fit into the computation model.
>>
>> Actually, each input is a fresh computation, with the previous
>> conversation history as input.
>>
>
> That seems correct. None-the-less it does get an increasingly
> deeper understanding of the subject matter, thus is not at
> all "deterministic" in any conventional use of the term.

On 2024-02-20 00:10:54 +0000, immibis said:

> On 19/02/24 16:46, olcott wrote:
>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>> On 2/19/24 1:03 AM, olcott wrote:
>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>
>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor infinite.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>> no
>>>>>>>>>>>>
>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>
>>>>>>>>>>> What would you expect someone who was paying attention to see?
>>>>>>>>>>
>>>>>>>>>> There is such a thing as incorrect yes/no questions that must
>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>
>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>
>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>
>>>>>>>> "This sentence is not true."
>>>>>>>>
>>>>>>>> *Increasing specificity*
>>>>>>>> (1) Says something about something.
>>>>>>>> (2) Says something about some sentence.
>>>>>>>> (3) Says that something is not true.
>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>> (5) Says that itself is not true.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>
>>>>>>> "Does this computation halt?"
>>>>>>>
>>>>>>> are precise questions which are not vague.
>>>>>>
>>>>>> A human can see that this is the question:
>>>>>> Do you halt on your own Turing Machine Description?
>>>>>
>>>>> A Turing machine has no way to access its own description.
>>>>
>>>> No one has ever correctly shown that a human mind is strictly stronger
>>>> than a TM. Therefore a TM can determine that it is processing its own
>>>> description.
>>>
>>> OBviosly you haven't been reading.
>>>
>>> The Human mind is not limited by the deterministic requirement, or the
>>> requirement to only use the designated input.
>>>
>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>
>>> Totally different.
>>>
>>> Maybe YOU have lost you ability to independently think, having traded
>>> your free-will to the devil for something, and are now enslaved on a
>>> predetermined path that you are stuck on forever now.
>>
>> Show me a technical paper that attempts to prove that the
>> human mind can perform computations that will remains forever
>> impossible for any advance in technological innovation of
>> computations.
>>
>> LLMs have already broken through any limitation of determinism.
>
> LLMs are deterministic, actually, unless you connect them to a random
> number generator. In fact, a random number generator is an essential
> part of the algorithm that generates text using a GPT model. If you
> change the random number generator with a non-random number generator,
> that always returns 0.4 for example, then it will be deterministic.

When determinism is not essential to the purpose of the device it is easier
to take than not to take an extra input from the environment and use that
as a random number.

For many pruposes a deterministic pseudorandom generator is as good as a
real random number generator.

--
Mikko

On 2/20/2024 6:34 AM, Richard Damon wrote:
> On 2/19/24 11:09 PM, olcott wrote:
>> On 2/19/2024 10:00 PM, immibis wrote:
>>> On 20/02/24 04:47, Richard Damon wrote:
>>>> On 2/19/24 7:19 PM, olcott wrote:
>>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> What would you expect someone who was paying attention
>>>>>>>>>>>>>>>> to see?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that
>>>>>>>>>>>>>>> must
>>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>>
>>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>>
>>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>>
>>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>>
>>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>>
>>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>>
>>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>
>>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>>
>>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>>> stronger
>>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>>> its own
>>>>>>>>> description.
>>>>>>>>
>>>>>>>> OBviosly you haven't been reading.
>>>>>>>>
>>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>>> or the requirement to only use the designated input.
>>>>>>>>
>>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>>
>>>>>>>> Totally different.
>>>>>>>>
>>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>>>
>>>>>>> Show me a technical paper that attempts to prove that the
>>>>>>> human mind can perform computations that will remains forever
>>>>>>> impossible for any advance in technological innovation of
>>>>>>> computations.
>>>>>>>
>>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>>
>>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>>> random number generator. In fact, a random number generator is an
>>>>>> essential part of the algorithm that generates text using a GPT
>>>>>> model. If you change the random number generator with a non-random
>>>>>> number generator, that always returns 0.4 for example, then it
>>>>>> will be deterministic.
>>>>>
>>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>>
>>>>
>>>> They might respond differently then they did before, but that is
>>>> because each "input" isn't a "fresh" computation, but a continuation
>>>> of the previous one.
>>>>
>>>> "Conversations" don't really fit into the computation model.
>>>
>>> Actually, each input is a fresh computation, with the previous
>>> conversation history as input.
>>>
>>
>> That seems correct. None-the-less it does get an increasingly
>> deeper understanding of the subject matter, thus is not at
>> all "deterministic" in any conventional use of the term.
>
> So, you don't knoiw the meaning of "Deterministic".
>
> If you restart the AI with the same state (including random number
> seeds) then it will answer the same prompt with the same response, EVERY
> TIME.
>

On 20/02/24 13:34, Richard Damon wrote:
> On 2/19/24 11:09 PM, olcott wrote:
>> On 2/19/2024 10:00 PM, immibis wrote:
>>> On 20/02/24 04:47, Richard Damon wrote:
>>>> On 2/19/24 7:19 PM, olcott wrote:
>>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> What would you expect someone who was paying attention
>>>>>>>>>>>>>>>> to see?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that
>>>>>>>>>>>>>>> must
>>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>>
>>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>>
>>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>>
>>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>>
>>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>>
>>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>>
>>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>
>>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>>
>>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>>> stronger
>>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>>> its own
>>>>>>>>> description.
>>>>>>>>
>>>>>>>> OBviosly you haven't been reading.
>>>>>>>>
>>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>>> or the requirement to only use the designated input.
>>>>>>>>
>>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>>
>>>>>>>> Totally different.
>>>>>>>>
>>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>>> enslaved on a predetermined path that you are stuck on forever now.
>>>>>>>
>>>>>>> Show me a technical paper that attempts to prove that the
>>>>>>> human mind can perform computations that will remains forever
>>>>>>> impossible for any advance in technological innovation of
>>>>>>> computations.
>>>>>>>
>>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>>
>>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>>> random number generator. In fact, a random number generator is an
>>>>>> essential part of the algorithm that generates text using a GPT
>>>>>> model. If you change the random number generator with a non-random
>>>>>> number generator, that always returns 0.4 for example, then it
>>>>>> will be deterministic.
>>>>>
>>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>>
>>>>
>>>> They might respond differently then they did before, but that is
>>>> because each "input" isn't a "fresh" computation, but a continuation
>>>> of the previous one.
>>>>
>>>> "Conversations" don't really fit into the computation model.
>>>
>>> Actually, each input is a fresh computation, with the previous
>>> conversation history as input.
>>>
>>
>> That seems correct. None-the-less it does get an increasingly
>> deeper understanding of the subject matter, thus is not at
>> all "deterministic" in any conventional use of the term.
>
> So, you don't knoiw the meaning of "Deterministic".
>
> If you restart the AI with the same state (including random number
> seeds) then it will answer the same prompt with the same response, EVERY
> TIME.
>
> The "learning" you are seeing is the evolution of the state of the
> neuron bank, so changing the input to the machine each time.

On 20/02/24 14:59, olcott wrote:
> On 2/20/2024 6:34 AM, Richard Damon wrote:
>> On 2/19/24 11:09 PM, olcott wrote:
>>> On 2/19/2024 10:00 PM, immibis wrote:
>>>> On 20/02/24 04:47, Richard Damon wrote:
>>>>> On 2/19/24 7:19 PM, olcott wrote:
>>>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> What would you expect someone who was paying attention
>>>>>>>>>>>>>>>>> to see?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that
>>>>>>>>>>>>>>>> must
>>>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>>>
>>>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>>>
>>>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>>>
>>>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>
>>>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>>>
>>>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>>>> stronger
>>>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>>>> its own
>>>>>>>>>> description.
>>>>>>>>>
>>>>>>>>> OBviosly you haven't been reading.
>>>>>>>>>
>>>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>>>> or the requirement to only use the designated input.
>>>>>>>>>
>>>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>>>
>>>>>>>>> Totally different.
>>>>>>>>>
>>>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>>>> enslaved on a predetermined path that you are stuck on forever
>>>>>>>>> now.
>>>>>>>>
>>>>>>>> Show me a technical paper that attempts to prove that the
>>>>>>>> human mind can perform computations that will remains forever
>>>>>>>> impossible for any advance in technological innovation of
>>>>>>>> computations.
>>>>>>>>
>>>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>>>
>>>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>>>> random number generator. In fact, a random number generator is an
>>>>>>> essential part of the algorithm that generates text using a GPT
>>>>>>> model. If you change the random number generator with a
>>>>>>> non-random number generator, that always returns 0.4 for example,
>>>>>>> then it will be deterministic.
>>>>>>
>>>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>>>
>>>>>
>>>>> They might respond differently then they did before, but that is
>>>>> because each "input" isn't a "fresh" computation, but a
>>>>> continuation of the previous one.
>>>>>
>>>>> "Conversations" don't really fit into the computation model.
>>>>
>>>> Actually, each input is a fresh computation, with the previous
>>>> conversation history as input.
>>>>
>>>
>>> That seems correct. None-the-less it does get an increasingly
>>> deeper understanding of the subject matter, thus is not at
>>> all "deterministic" in any conventional use of the term.
>>
>> So, you don't knoiw the meaning of "Deterministic".
>>
>> If you restart the AI with the same state (including random number
>> seeds) then it will answer the same prompt with the same response,
>> EVERY TIME.
>>
>
> Not necessarily. It is currently not configured to update its master
> knowledgebase after every conversation. It could be configured to update
> its master knowledgebase after every interaction. It does update its
> discourse knowledgebase after every interaction.

On 20/02/24 12:56, Ben Bacarisse wrote:
> immibis <news@immibis.com> writes:
>
>> The subject is whether D(D) has a finite or infinite execution
>> sequence.
>
> I case it helps you decide the value of the interactions you are having
> (I know it can be fun) PO has stated categorically that the wrong answer
> is the right one:
>
> Me: Here's the key question: do you still assert that H(P,P) == false is
> the "correct" answer even though P(P) halts?
>
> PO: Yes that is the correct answer even though P(P) halts
I know this one. It is based on an incorrect understanding of a lemma.
Using the incorrect version of the lemma he "proves" that false is the
correct answer (even though it isn't).

> And, so you know what you might be getting into he is prepared to say
> things like this:
>
> "the fact that a computation halts does not entail that it is a
> halting computation"
>

Indeed. Quite funny.

> and, when pressed, he is even prepared to state categorical opposites
> and not correct either:
>
> "Furthermore I have repeated H.q0 ⟨Ĥ⟩ ⟨Ĥ⟩ transitions to H.qn many
> times."
>
> "No nitwit H ⟨Ĥ⟩ ⟨Ĥ⟩ transitions to H.qy as I have told you many
> times."
>

I don't remember that one. Also quite funny.

On 2/20/24 8:59 AM, olcott wrote:
> On 2/20/2024 6:34 AM, Richard Damon wrote:
>> On 2/19/24 11:09 PM, olcott wrote:
>>> On 2/19/2024 10:00 PM, immibis wrote:
>>>> On 20/02/24 04:47, Richard Damon wrote:
>>>>> On 2/19/24 7:19 PM, olcott wrote:
>>>>>> On 2/19/2024 6:10 PM, immibis wrote:
>>>>>>> On 19/02/24 16:46, olcott wrote:
>>>>>>>> On 2/19/2024 6:49 AM, Richard Damon wrote:
>>>>>>>>> On 2/19/24 1:03 AM, olcott wrote:
>>>>>>>>>> On 2/18/2024 8:55 PM, immibis wrote:
>>>>>>>>>>> On 19/02/24 01:13, olcott wrote:
>>>>>>>>>>>> On 2/18/2024 6:00 PM, immibis wrote:
>>>>>>>>>>>>> On 19/02/24 00:47, olcott wrote:
>>>>>>>>>>>>>> On 2/18/2024 11:36 AM, Richard Damon wrote:
>>>>>>>>>>>>>>> On 2/18/24 10:33 AM, olcott wrote:
>>>>>>>>>>>>>>>> On 2/18/2024 4:57 AM, Mikko wrote:
>>>>>>>>>>>>>>>>> On 2024-02-18 01:40:47 +0000, olcott said:
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> On 2/17/2024 4:26 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>> On 17/02/24 16:01, olcott wrote:
>>>>>>>>>>>>>>>>>>>> On 2/17/2024 4:19 AM, Mikko wrote:
>>>>>>>>>>>>>>>>>>>>> On 2024-02-16 22:06:41 +0000, olcott said:
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> On 2/16/2024 3:55 PM, immibis wrote:
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>>> Show me a sequence that is neither finite nor
>>>>>>>>>>>>>>>>>>>>>>> infinite.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>>> It seems that you still can't grasp the notion of
>>>>>>>>>>>>>>>>>>>>>> yes/no questions having no correct yes/no answer.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>> It seems that Olcott doesn't understand that there
>>>>>>>>>>>>>>>>>>>>> are questions that are not yes/no question.
>>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>> My primary point (since 2004) is that any yes/no
>>>>>>>>>>>>>>>>>>>> question defined to have no correct yes/no answer
>>>>>>>>>>>>>>>>>>>> is an incorrect question.
>>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>>> Can Olcott compute the square root of a pickle?
>>>>>>>>>>>>>>>>>>> is an incorrect question
>>>>>>>>>>>>>>>>>>> but the answer is:
>>>>>>>>>>>>>>>>>>> no
>>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>> You just flat out are not paying attention.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> What would you expect someone who was paying attention
>>>>>>>>>>>>>>>>> to see?
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> There is such a thing as incorrect yes/no questions that
>>>>>>>>>>>>>>>> must
>>>>>>>>>>>>>>>> be rejected as semantically unsound because they have been
>>>>>>>>>>>>>>>> defined such that both yes and no are the wrong answer.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Bu, "Does the Computation described by this input Halt?"
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *IS MORE VAGUE THAN*
>>>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> "This sentence is not true."
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> *Increasing specificity*
>>>>>>>>>>>>>> (1) Says something about something.
>>>>>>>>>>>>>> (2) Says something about some sentence.
>>>>>>>>>>>>>> (3) Says that something is not true.
>>>>>>>>>>>>>> (4) Says that some sentence is not true.
>>>>>>>>>>>>>> (5) Says that itself is not true.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> It is irrelevant. ALL questions of the form:
>>>>>>>>>>>>>
>>>>>>>>>>>>> "Does this computation halt?"
>>>>>>>>>>>>>
>>>>>>>>>>>>> are precise questions which are not vague.
>>>>>>>>>>>>
>>>>>>>>>>>> A human can see that this is the question:
>>>>>>>>>>>> Do you halt on your own Turing Machine Description?
>>>>>>>>>>>
>>>>>>>>>>> A Turing machine has no way to access its own description.
>>>>>>>>>>
>>>>>>>>>> No one has ever correctly shown that a human mind is strictly
>>>>>>>>>> stronger
>>>>>>>>>> than a TM. Therefore a TM can determine that it is processing
>>>>>>>>>> its own
>>>>>>>>>> description.
>>>>>>>>>
>>>>>>>>> OBviosly you haven't been reading.
>>>>>>>>>
>>>>>>>>> The Human mind is not limited by the deterministic requirement,
>>>>>>>>> or the requirement to only use the designated input.
>>>>>>>>>
>>>>>>>>> Humans have "Free-Will", while Turing Macines are "Deterministic".
>>>>>>>>>
>>>>>>>>> Totally different.
>>>>>>>>>
>>>>>>>>> Maybe YOU have lost you ability to independently think, having
>>>>>>>>> traded your free-will to the devil for something, and are now
>>>>>>>>> enslaved on a predetermined path that you are stuck on forever
>>>>>>>>> now.
>>>>>>>>
>>>>>>>> Show me a technical paper that attempts to prove that the
>>>>>>>> human mind can perform computations that will remains forever
>>>>>>>> impossible for any advance in technological innovation of
>>>>>>>> computations.
>>>>>>>>
>>>>>>>> LLMs have already broken through any limitation of determinism.
>>>>>>>
>>>>>>> LLMs are deterministic, actually, unless you connect them to a
>>>>>>> random number generator. In fact, a random number generator is an
>>>>>>> essential part of the algorithm that generates text using a GPT
>>>>>>> model. If you change the random number generator with a
>>>>>>> non-random number generator, that always returns 0.4 for example,
>>>>>>> then it will be deterministic.
>>>>>>
>>>>>> LLMs can change their mind in the middle of a single dialogue.
>>>>>>
>>>>>
>>>>> They might respond differently then they did before, but that is
>>>>> because each "input" isn't a "fresh" computation, but a
>>>>> continuation of the previous one.
>>>>>
>>>>> "Conversations" don't really fit into the computation model.
>>>>
>>>> Actually, each input is a fresh computation, with the previous
>>>> conversation history as input.
>>>>
>>>
>>> That seems correct. None-the-less it does get an increasingly
>>> deeper understanding of the subject matter, thus is not at
>>> all "deterministic" in any conventional use of the term.
>>
>> So, you don't knoiw the meaning of "Deterministic".
>>
>> If you restart the AI with the same state (including random number
>> seeds) then it will answer the same prompt with the same response,
>> EVERY TIME.
>>
>
> Not necessarily. It is currently not configured to update its master
> knowledgebase after every conversation. It could be configured to update
> its master knowledgebase after every interaction. It does update its
> discourse knowledgebase after every interaction.