Thoughts about Consciousness of AI
Are Generative AI Systems ChatGPT, Google Gemini, xAI Grok, Baidu Ernie, Bhashini, AppleGPT self-referential?
Question to ChatGPT: "chatgpt do you belong to generative AI?"
Answer from ChatGPT: "Yes I belong to Generative AI category."
Question to ChatGPT: "are you self-referential?"
Answer from ChatGPT: "ChatGPT is a prototype chatbot, which is a text-based dialog system as a user interface based on machine learning.
ChatGPT is not self-referential, as it does not make any statements about itself."
>>> Digression: ChatGPT on Luhmann's self-referential systems.
Niklas Luhmann was a German sociologist and founder of the sociological theory of self-referential systems. In his theory construction he followed the work of Talcott Parsons and wanted to found a subject-universal theory, which corresponds in its complexity content to the complexity of real social systems - up to social systems.
Self-reference occurs only in combination with other-reference.
Before we address the question
"Where can Generative AI replace Human Intelligence?"
we should deal with the following question in more detail:
Are Generative AI Systems self-referential?
About Self-Reference as Basis of our Concept "Auto-dynamical Systems"
The concept of self-reference describes the fact that there are systems that refer to themselves through their operations [→Operation / Observation].
These are (organic, psychic, social) systems that can only observe reality because of this self-contact [→System/Environment].
Self-reference occurs when the operation of observation is included in what is referred to, i.e., when it refers to something that belongs to it.
A social system, for instance, can only reproduce communication and can only take the world into account through communication; self-reference is implied in every →communication in the form of utterance.
In the same way, a consciousness can only think and reality can only gain relevance to the consciousness as the external referent of thought. The concept of self-reference is intended to be neither purely analytical, nor a characteristic of the transcendental subject.
System theory observes neither the human being nor the subject.
The self-referential constitution of the organic, psychic and social systems is recognized as an empirical finding: such systems exist in the real world and are really self-referential.
This description attempts to overcome the conflict between nominalism and realism; self-reference does not describe a solipsistic or transcendental conception of the world.
Systems constituted self-referentially must be able to distinguish between what belongs to the system (its own operations) and what is to be attributed to its environment.
On the one hand, self-reference requires the possibility to reproduce the system’s own operations [→Autopoiesis] in such a way that every distinction used for observation (i.e., for the description of something) must be construed by the operations within the system itself.
On the other hand, the system cannot mingle with the external reality, i.e., with its environment; the condition for its operativity and for each form of cognition is the possibility to distinguish internally between self- and other-reference.
This ability distinguishes self-referential systems from trivial machines according to One and Only Heinz von Foerster :
while the latter always transform inputs into outputs in the same way, for self-referential machines the output depends on the particular internal state of the system; thus, depending on the different states of the system, the same input can lead to completely different results.
The concept of self-reference is not interchangeable with that of tautology. It does not concern an operation that directly describes itself (for example A=A), but rather an operation that refers to something (the “self ” of the reference) that belongs to it.
This reference is only possible because a difference which allows something that refers to itself is differentiated from something else.
A tautology would be a non-informative form of self-reference, which would be fatal for the operations of the system: excluding the reference to other things would block every operative connection [→Asymmetrization].
Depending on which difference is used to refer to the “self ”, self-reference can be specified differently:
(a) When what is referring to itself is an element of the system (a communication, a thought, a cell) we speak of basal self-reference.
The concept of the element refers to a fundamental unit of the system which cannot be further decomposed.
Firstly, every element is an element only in reference to a system, and no elements exist without the system within which the elements are elements.
Secondly, elements exist only in relation to other elements; what constitutes them is the difference between element and relation, and their behavior. The element can refer to itself because of the differentiation of element and relation.
This differentiation is the basis upon which the system operations make circular references to themselves within connections to other operations.
Self-reference is the form of meaning, since what is actualized returns to itself within the reference to the possible.
In the case of social systems, there is basal self-reference insofar as communications have no other points of reference than other communications:
only due to this reference do they allow the autopoiesis of the system.
The relations permitted between the elements are selected by the structures of the system. In this sense, the difference between element and relation cannot be observed at the level of structures, but only at the level of autopoiesis.
Meaning-constituting systems produce their own elements as operations and they have thorough responsibility for this production: the identity and the quality of an element can only be constituted within the system for which the element is an element. Input from operations from outside and the external definition of the relations between them are both unconceivable, because that would mean the destruction of the system.
(b) When the “self ” of the reference is a process, we speak of reflexivity that can be observed on the basis of the difference between before and after.
Reflexivity consists in strengthening the selectivity of the process through applying the process to itself, which takes priority over applying it to that which is processed.
An example is the learning to learn: instead of referring to the object of learning directly, it refers to the process of learning itself, such that its ability and its selectivity will be strengthened.
(c) In the third case the system refers to itself through its own operations, and this demands that the system can be differentiated internally (itself) and externally (its environment).
In this case, we speak of reflection.
The difference, which in this case guides self-reference, is that of system and environment. In reflection, a re-entry of the difference system/environment is realized within the system.
This is done with the help of a further operation in the system itself, and in this sense reflection continues the autopoiesis, but at the same time makes it possible for the system to gain information about itself.
This is conveyed by the system in the form of a difference (the difference system/environment), which it represents as a contingent entity and compares with alternative possibilities. In all these types of self-reference, we are dealing with a closed circularity that nevertheless does not deny the existence of the environment; on the contrary, the environment is a prerequisite for the system’s selections. Selfreferential systems are autonomous systems that use this closure to maintain their own autopoiesis and make their own observations possible.
For this reason, we should not understand autonomy as independence, but rather as selfreferential closure: the environment can limit or increase the range of operative possibilities, but this does not change the fact that the operations can only be produced and linked with one another by and in the system. For this reason, autonomy is always absolute and not relative, since it would not make sense to conceive of a system that is only partly autonomous, or only closed “a bit”. External, observable influences on the system only affect its degree of irritability or the performance demands made by other systems, never its autonomy or closure.
Thus autonomy should be understood as the relationship of dependence and independence between system and environment, where this difference can only be drawn within the system, i.e., self-referentially.
For instance, scientific research surely depends on the financial resources (economic operations) available, but these resources cannot buy truth. The scientific system can increase its own complexity and structure itself accordingly if it becomes less and less susceptible to the influence of moral, religious or political demands.
If, however, science has become a subsystem of the functionally differentiated society, it alone can make decisions regarding the production of knowledge, research and the difference between true and untrue.
We can thus observe a system which reproduces itself at an autopoietic level through its own basal self-reference, and develops its own reflection in the form of the theory of knowledge.
The relevance of the environment is not denied by the self-referential constitution of this system: we conceive of the environment as a complexity that can be determined only by the structures of the system.
However, it is the complexity of the environment that allows a system-internal complexity to be constructed.
The concept of self-reference therefore rules out any continuity of system and environment.
This implies that every description of the environment through the system (i.e., every other-reference and every opening) is only possible as a construction of the system.
But self-reference is not to be equated with the observer, even if at first glance they look the same. The form of the observer does not lie in mere self-description, but in the difference between self- and other-reference as a difference.
The observer is the unity of this difference:
the self can only be mentioned in contrast to the other, and in general the unity of the observer appears paradoxical, in that it is the unity of the difference, the simultaneity of self- and other-reference.
(Source: Unlocking Luhmann by Claudio Baraldi, Giancarlo Corsi, Elena Esposito)
"Nothing in sociology makes sense except in the light of game theory" - Pedro Domingos, Professor of computer science
Picture: Payoff Matrix for Artificial Intelligence in Mediocristan und Extremistan
Where can Generative AI replace Human Intelligence?
- Fragility or
- Resilience (Robustness) or
- Antifragility or
Fragility or antifragility refers to the potential benefit or disadvantage of being exposed to something related to volatility.
These something are
- Imperfect, incomplete knowledge
- The Unknown
- Dispersion of Results
Notice: Uncertainty, Disorder, and The Unknown can have the same effect:
Antifragile systems benefit to some degree.
Fragile systems suffer from it.
(Source: Antifragility by Nassim Nicholas Taleb)
Picture: Payoff Matrix for Machine Learning in Mediocristan und Extremistan
Picture: Generative AI - ChatGPT, Google BARD, Baidu ERNIE: Optimizing Language Models for Dialogue in Functional Systems of Society
Picture: Generative AI - ChatGPT, Google BARD, Baidu ERNIE: Optimizing Language Models for Dialogue in between Functional Systems of Society
About Structural Coupling between Functional Systems
"In a functionally differentiated society, structural coupling presupposes the operational closure of functional systems.
However, structural coupling between functional systems is also operational coupling, because specific communications are contingently shared by different systems, although they are immediately connected to the internal autopoiesis of these systems (e.g., to the political system and the economic system or the legal system and the economic system).
For instance, negotiating and signing a property contract is a communication which is contingently shared by the legal system and the economic system, and the government’s reimbursement of tax is a communication which is contingently shared by the political system and the economic system.
Operational coupling is possible in society because it involves the same type of operation (communication), but it is not possible between social systems and psychic systems, as communication and thought are different types of operation, ensuring the autopoiesis of different types of meaning-constituting systems.
Structural and operational coupling between functional systems show how functional differentiation determines the integration of society.
This integration is necessary because the differentiation of functions also requires systematic interdependencies to reproduce society as a whole.
However, the great quantity of structural and operational couplings between the subsystems exposes the functionally differentiated society to continuous and systematic self-irritations, which cannot be regulated at the level of the society itself. Therefore, in the functionally differentiated society, the combination of multiple forms of operational closure in functional systems, and the interdependencies that enhance continuous self-irritation in each functional system, create an exceptional level of complexity."
(Source: Unlocking Luhmann by Claudio Baraldi, Giancarlo Corsi, Elena Esposito)