The Conscious Turing Machine

105 views
Skip to first unread message

Philip Thrift

unread,
Nov 24, 2020, 2:17:34 PM11/24/20
to Everything List
The Conscious Turing Machine

video:

arXiv:

A Theoretical Computer Science Perspective on Consciousness

The quest to understand consciousness, once the purview of philosophers and theologians, is now actively pursued by scientists of many stripes. This paper studies consciousness from the perspective of theoretical computer science. It formalizes the Global Workspace Theory (GWT) originated by cognitive neuroscientist Bernard Baars and further developed by him, Stanislas Dehaene, and others. Our major contribution lies in the precise formal definition of a Conscious Turing Machine (CTM), also called a Conscious AI. We define the CTM in the spirit of Alan Turing's simple yet powerful definition of a computer, the Turing Machine (TM). We are not looking for a complex model of the brain nor of cognition but for a simple model of (the admittedly complex concept of) consciousness. After formally defining CTM, we give a formal definition of consciousness in CTM. We then suggest why the CTM has the feeling of consciousness. The reasonableness of the definitions and explanations can be judged by how well they agree with commonly accepted intuitive concepts of human consciousness, the breadth of related concepts that the model explains easily and naturally, and the extent of its agreement with scientific evidence.




@philipthrift

Telmo Menezes

unread,
Nov 25, 2020, 7:05:02 AM11/25/20
to Philip Thrift, 'Brent Meeker' via Everything List
A long time ago I read a self-help book (don't judge me) that included a story that goes something like this:

A workers team is cleaning a forest. A wise guy climbs the tallest tree and shouts: "wrong forest!". The manager replies: "shut up, we're making progress!".

Telmo
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.

John Clark

unread,
Nov 25, 2020, 9:58:14 AM11/25/20
to everyth...@googlegroups.com


On Tue, Nov 24, 2020 at 2:17 PM Philip Thrift <cloud...@gmail.com> wrote:

> The Conscious Turing Machine
arXiv:

DEFINITION 1.5.1. At each time t ≥ 0, Short Term Memory holds exactly one chunk, which is designated the entirety of Conscious Turing Machine's conscious content. Consciousness in Conscious Turing Machine is defined to be the awareness (i.e. the reception by all Long Term Memory). It defines the conscious content of Conscious Turing Machine to be whatever chunk is in Short Term Memory; and then defines Conscious Awareness by the Conscious Turing Machineto be the reception by all Long Term Memory processors of Short Term Memory's broadcast of that content. The gists of those broadcasts may be viewed as the inner thoughts generated by Conscious Turing Machine's unconscious processors or the speech, vision, touch, taste, and/or whatever else is received as input by the Conscious Turing Machine

Consciousness and awareness mean the same thing, so all the verbiage in the above could be boiled down to "Consciousness in Conscious Turing Machine theory is defined to be consciousness" which I don't find to be terribly helpful. And in the article I don't see even a hint as to how an experimental test could be performed to see if this "theory" is actually correct.

John K Clark

Bruno Marchal

unread,
Nov 25, 2020, 11:00:11 AM11/25/20
to everyth...@googlegroups.com
Manuel Blum and his wife Lenore Blum wrote many quite interesting paper in theoretical computer, based on the work of Gold on "limiting recursion” and which inspired also the very interesting work of Case & Smith. 

But here, they are naïve on consciousness and they are still reasoning in an Aristotelian Metaphysics, with the usual identity thesis which I have shown to be incompatible with Mechanism.

Once you work in the digital mechanist frame, a theory of consciousness is offered by the necessary first person mode (given by the consistency of truth nuance added to provability (rational belief). Incompleteness simply does not leave any choice: the logic of p (sigma_1), []p (sigma_1-complete), []p & p (the subject, as we know by thought experience, or by using Theaetus’definition), []p & <>t, []p & <>t & p, obeys different mathematics, explaining why the arithmetical reality when seen from inside takes the psycho-theological shapes, including the physical reality.

This predicted quickly the “many-histories’ aspect of nature, and has been confirmed by attributing quantum logics to the platonic material modes.

Incompleteness explains both 1) the vertical nuances between, truth, belief, knowledge, observation, and sensation:

P
[]p
[]p & p
[]p & <>t
[]p & <>t & p,

 and 2) the horizontal nuance between provable and true which splits 3 modes in two, leading to the 8 arithmetical “hypostases”. That difference explains (mathematically) where the qualia and quanta comes from, and how they are related, and why it looks like, locally, the identity thesis is true, and why it works (locally). 

Aristotle confused the “apparent physical universe” with an ontological universe. There are no evidences at all for the first one, albeit a lot of evidence for the second one, including an explanation of how the illusion is unavoidable for most universal machine (in arithmetic).

Mechanism changes the burden of the proof. A Materialist has to abandon Digital Mechanism (and thus Darwin!), and has to provide some evidence for an ontological universe, and some evidences that it differs from the “canonical physics of the universal (reflexive, Löbian) machine/number. Up to now, no one has found a discrepancy…

Consciousness is simply knowledge, and incompleteness assures it exists on its own, and has all the platonic characteristic intuited in Parmenides, Moderatus, Plotinus, Proclus, etc… (unfortunately not well known).

And physics is explained, both in its qualia sensible and non sharable aspects, as well as its quanta, at least up to now.

In a sense, Mechanism enforces a sort of strong Aristotelian Atheism, as there is no creator, nor a creation, just a universal dreamer, which can awake from time to time. But, from a Platonic perspective, it needs at the least a Pythagorean ontology (numbers), as it necessitates to postulate at least one universal machinery (like very elementary arithmetic (Robinson), or combinators, etc.).

It can be proved that we cannot prove/explain/define the existence of a universal machinery from anything less.

Blum and Blum are probably interesting, but incomplete and unaware of the canonical “theory of everything” provided already by the universal machine or number. They are still avoiding the “hard” part of the problem: the mind-body problem, and they use in the background the “wrong” ontology (wrong with Mechanism).

The reflexive universal machine knows that is true and provable for simpler (than itself) reflexive universal machine, and they know that they have a soul, and that such a theory is true if classical mechanism is true, which they know they cannot prove, unless inconsistent.

Bruno

Brent Meeker

unread,
Nov 25, 2020, 3:59:43 PM11/25/20
to everyth...@googlegroups.com
"Charles" Dennett?  Any relation to Daniel?

Brent

Lawrence Crowell

unread,
Nov 25, 2020, 5:52:39 PM11/25/20
to Everything List
Blum defined a funny sort of meaning to incomputability, which is basically recursively enumerable. Blum stated this was incomputable because the computation would not halt and there is no final output. This differs from Turing's definition in that one cannot determine if all things halt or do not halt. 

The video is sort of long and I am not sure if I have the time to devote to this. This definition of inner thought being conferred to a system which exhibits the outward properties of a person known to have inner thoughts is not much different from the Turing test definition. This is couched in the definition of memory or long term memory as a cache of short term memories. This seems rather artificial to me.

LC 

Philip Thrift

unread,
Nov 26, 2020, 6:59:53 AM11/26/20
to Everything List
I am unconscious of where they where they said "Charles".

@philipthrift

Philip Thrift

unread,
Nov 26, 2020, 7:28:53 AM11/26/20
to Everything List
The architecture they specify in their paper


could be of interest to the AI hacker (MIT AI Lab type) trying to implement and demonstrate a "conscious" robot" (even just an impressively "fake" one). 

We agree with Christof Koch that “There isn’t a Turing Test for consciousness. You have to look at the way the system is built. You have to look at the circuitry, not [only] its behavior”. We would emphasize “architecture” over “circuitry”.

(But the "neuromorphic" might not just be architecture, but material composition.)

It is interesting that the presence of these artificial beings may be looked at differently depending on culture.

Why Westerners Fear Robots and the Japanese Do Not

@philipthrift

Lawrence Crowell

unread,
Nov 26, 2020, 9:17:58 AM11/26/20
to Everything List
I have my concerns not so much with robots, but with their development in various directions. An example is this:


LC

PGC

unread,
Nov 26, 2020, 9:26:02 AM11/26/20
to Everything List
On Thursday, November 26, 2020 at 1:28:53 PM UTC+1 cloud...@gmail.com wrote:
The architecture they specify in their paper


could be of interest to the AI hacker (MIT AI Lab type) trying to implement and demonstrate a "conscious" robot" (even just an impressively "fake" one). 

We agree with Christof Koch that “There isn’t a Turing Test for consciousness. You have to look at the way the system is built. You have to look at the circuitry, not [only] its behavior”. We would emphasize “architecture” over “circuitry”.

Show me a set of systems that demonstrate indeterminacy between pursuing some choice of goal(s) and doing nothing. If they demonstrate activity wherein a goal is repurposed or glitched as an end in itself; i.e. where the goal's completion/felicity becomes irrelevant to the continued pursuit of the activity, which itself demonstrates constant refinement and increased specificity at each iteration as in social interaction, sex, drugs, rock and roll etc. you're closer to a conscious machine, regardless of architecture, circuitry, way the system is built etc. Particularly if it asks the admin for more resources to pursue its ends. One difficulty lies in demonstrating that the system(s) achieved such freely; hence you have to show that some significant number of them did nothing and/or other things. PGC

John Clark

unread,
Nov 26, 2020, 12:56:52 PM11/26/20
to everyth...@googlegroups.com
On Thu, Nov 26, 2020 at 7:28 AM Philip Thrift <cloud...@gmail.com> wrote:

> We agree with Christof Koch that “There isn’t a Turing Test for consciousness. You have to look at the way the system is built. You have to look at the circuitry, not [only] its behavior”. We would emphasize “architecture” over “circuitry”.

For many thousands of years people have assumed that their fellow human beings were conscious when they were behaving intelligently but were not conscious when they weren't, such as when they were sleeping or dead; and yet they knew absolutely nothing about the circuitry or the architecture of anybody's brain or even knew that the brain had something to do with thinking. Were they making a mistaken assumption for all these millennia, should they have adopted a more solipsistic attitude?

John K Clark

Brent Meeker

unread,
Nov 26, 2020, 3:15:53 PM11/26/20
to everyth...@googlegroups.com


On 11/26/2020 4:28 AM, Philip Thrift wrote:
The architecture they specify in their paper


could be of interest to the AI hacker (MIT AI Lab type) trying to implement and demonstrate a "conscious" robot" (even just an impressively "fake" one). 

We agree with Christof Koch that “There isn’t a Turing Test for consciousness. You have to look at the way the system is built. You have to look at the circuitry, not [only] its behavior”. We would emphasize “architecture” over “circuitry”.

Then can they answer the question, "Is an octopus conscious?"  Octopuses have a very different neural architecture.  Two thirds of their neurons are in their legs.  They have almost as many as a dog (5e8 vs 7e8).  They play.  They recognize individual people.  I'd say they are conscious...but then I'm only looking at behavior.  I curious as to what looking at architecture would contribute?

Brent


(But the "neuromorphic" might not just be architecture, but material composition.)

It is interesting that the presence of these artificial beings may be looked at differently depending on culture.

Why Westerners Fear Robots and the Japanese Do Not

@philipthrift

On Wednesday, November 25, 2020 at 4:52:39 PM UTC-6 Lawrence Crowell wrote:
On Wednesday, November 25, 2020 at 8:58:14 AM UTC-6 johnk...@gmail.com wrote:


On Tue, Nov 24, 2020 at 2:17 PM Philip Thrift <cloud...@gmail.com> wrote:

> The Conscious Turing Machine
arXiv:

DEFINITION 1.5.1. At each time t ≥ 0, Short Term Memory holds exactly one chunk, which is designated the entirety of Conscious Turing Machine's conscious content. Consciousness in Conscious Turing Machine is defined to be the awareness (i.e. the reception by all Long Term Memory). It defines the conscious content of Conscious Turing Machine to be whatever chunk is in Short Term Memory; and then defines Conscious Awareness by the Conscious Turing Machineto be the reception by all Long Term Memory processors of Short Term Memory's broadcast of that content. The gists of those broadcasts may be viewed as the inner thoughts generated by Conscious Turing Machine's unconscious processors or the speech, vision, touch, taste, and/or whatever else is received as input by the Conscious Turing Machine

Consciousness and awareness mean the same thing, so all the verbiage in the above could be boiled down to "Consciousness in Conscious Turing Machine theory is defined to be consciousness" which I don't find to be terribly helpful. And in the article I don't see even a hint as to how an experimental test could be performed to see if this "theory" is actually correct.

John K Clark

Blum defined a funny sort of meaning to incomputability, which is basically recursively enumerable. Blum stated this was incomputable because the computation would not halt and there is no final output. This differs from Turing's definition in that one cannot determine if all things halt or do not halt. 

The video is sort of long and I am not sure if I have the time to devote to this. This definition of inner thought being conferred to a system which exhibits the outward properties of a person known to have inner thoughts is not much different from the Turing test definition. This is couched in the definition of memory or long term memory as a cache of short term memories. This seems rather artificial to me.

LC 
--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.

Philip Thrift

unread,
Nov 26, 2020, 4:57:03 PM11/26/20
to Everything List
On Thursday, November 26, 2020 at 2:15:53 PM UTC-6 Brent wrote:


On 11/26/2020 4:28 AM, Philip Thrift wrote:
The architecture they specify in their paper


could be of interest to the AI hacker (MIT AI Lab type) trying to implement and demonstrate a "conscious" robot" (even just an impressively "fake" one). 

We agree with Christof Koch that “There isn’t a Turing Test for consciousness. You have to look at the way the system is built. You have to look at the circuitry, not [only] its behavior”. We would emphasize “architecture” over “circuitry”.

Then can they answer the question, "Is an octopus conscious?"  Octopuses have a very different neural architecture.  Two thirds of their neurons are in their legs.  They have almost as many as a dog (5e8 vs 7e8).  They play.  They recognize individual people.  I'd say they are conscious...but then I'm only looking at behavior.  I curious as to what looking at architecture would contribute?

Brent

Their "architecture" is in the sense of a programming architecture (as would be needed in a "conscious" robot).


We see a kinship between the CTM and the self-aware robots developed by (Chella, Pipitone, Morin, & Racy, 2020).

(The cognitive architecture for inner and private speech)


@philipthrift

Bruno Marchal

unread,
Nov 27, 2020, 4:53:49 AM11/27/20
to everyth...@googlegroups.com
On 25 Nov 2020, at 23:52, Lawrence Crowell <goldenfield...@gmail.com> wrote:

On Wednesday, November 25, 2020 at 8:58:14 AM UTC-6 johnk...@gmail.com wrote:


On Tue, Nov 24, 2020 at 2:17 PM Philip Thrift <cloud...@gmail.com> wrote:

> The Conscious Turing Machine
arXiv:

DEFINITION 1.5.1. At each time t ≥ 0, Short Term Memory holds exactly one chunk, which is designated the entirety of Conscious Turing Machine's conscious content. Consciousness in Conscious Turing Machine is defined to be the awareness (i.e. the reception by all Long Term Memory). It defines the conscious content of Conscious Turing Machine to be whatever chunk is in Short Term Memory; and then defines Conscious Awareness by the Conscious Turing Machineto be the reception by all Long Term Memory processors of Short Term Memory's broadcast of that content. The gists of those broadcasts may be viewed as the inner thoughts generated by Conscious Turing Machine's unconscious processors or the speech, vision, touch, taste, and/or whatever else is received as input by the Conscious Turing Machine

Consciousness and awareness mean the same thing, so all the verbiage in the above could be boiled down to "Consciousness in Conscious Turing Machine theory is defined to be consciousness" which I don't find to be terribly helpful. And in the article I don't see even a hint as to how an experimental test could be performed to see if this "theory" is actually correct.

John K Clark

Blum defined a funny sort of meaning to incomputability, which is basically recursively enumerable.

This is just the usual Turing one, although expressed in a slightly ambiguous way. “Computable” means usually “total computable”.


Blum stated this was incomputable because the computation would not halt and there is no final output. This differs from Turing's definition in that one cannot determine if all things halt or do not halt. 


Halting is “partial computable”. A function is said non computable everywhere it is not defined, and that is itself only partially computable. Partially computable entails partially not computable in case the function is not total.



The video is sort of long and I am not sure if I have the time to devote to this. This definition of inner thought being conferred to a system which exhibits the outward properties of a person known to have inner thoughts is not much different from the Turing test definition.

Yes, and that is why it useless for the mind-body problem. 


This is couched in the definition of memory or long term memory as a cache of short term memories. This seems rather artificial to me.

It makes sense for some application, but avoid the usual difficulties, like the whole Aristotelian (weak-materialist) tradition. 

Bruno



LC 

--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.

Bruno Marchal

unread,
Nov 27, 2020, 5:01:48 AM11/27/20
to everyth...@googlegroups.com

Trying to detect (or understand) consciousness from the architecture, is like trying to explain alpha of deep blue by looking how a gate is working. The architecture might be contingent, and has nothing to do with an explanation which (assuming mechanism) is of an higher order, like an algorithm, or like a full psychology or theology of the entities concerned.

Consciousness, like Truth, is not among the definable notion, although it is easy to metadefined it when we assume Mechanism. (In that case, consciousness is the belief in a reality, in case there is a reality (which we can never justify rationally).

This does not prevent us to bet our fellows are conscious, or that there is some reality “beyond my personal experience”. 

Bruno


--
You received this message because you are subscribed to the Google Groups "Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to everything-li...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages