Non-subjective artificial intelligence in the system of subject-object relations

Gribkov Andrei Armovich ORCID: 0000-0002-9734-105X Doctor of Technical Science Leading researcher; Scientific and production complex 'Technological Center' 124498, Russia, Moscow, Zelenograd, Shokina Square, 1, building 7 andarmo@yandex.ru Other publications by this author

Introduction

The active development in the field of information technology observed in recent decades has now prepared all the necessary conditions for the creation of artificial intelligence systems. So far, artificial intelligence systems are represented by machine learning systems (the most advanced ones are deep machine learning systems) and have limited capabilities to solve intellectual problems. Nevertheless, the observed expansion of the scope of such systems and the growth of their capabilities are already actualizing a number of fundamental philosophical issues of great civilizational importance.

The function of artificial intelligence is to replace humans in solving intellectual tasks, the most important of which are tasks related to controlling machines that meet human needs. An intellectual task is called "for which there is no universally recognized (classical) solution, therefore, in order to solve such a problem, it is necessary to come up with a solution method" ^[1].

The question of what level artificial intelligence is capable of achieving does not yet have a definitive answer. One of the key factors determining the answer to this question is the presence or absence of the need for the subjectivity of artificial intelligence to bring its abilities closer to human ones. Subjectivity is understood as the ability to be a subject of activity, including in relation to objects of cognition. This understanding of subjectivity brings him closer to the self – the ability of the subject to get out from under the power of external causes and create a reproducible order of life, determined from within ^[2].

Previous studies ^[3] have shown that the answer to the question of the need for subjectivity depends on how we imagine the real world. If we imagine it as a non–deterministic system in which creativity is an irrational (inaccessible to comprehension by reason) process, then we cannot get a reliable answer to the question of the need for subjectivity. We only know that a person can solve intellectual problems and has subjectivity. If the world is deterministic, everything in it has reasons and (theoretically) can be explained, then there is no qualitative difference between a computational or logical task (which an ordinary computer copes with perfectly) and an intellectual task. In this case, subjectivity is not required for artificial intelligence to perform its function of solving intellectual tasks, but only the expansion of existing knowledge is necessary. A philosophical understanding of the primary properties of being and the structure of matter at the lowest level of its organization ^{[4; 5]} gives sufficient grounds to believe that in reality the world is not qualitatively infinitely complex, it is deterministic and it is based on mechanisms that have a rational explanation. Thus, it can be assumed with a high degree of confidence that subjectivity is not necessary to solve intellectual problems.

One of the defining hypostases of the theory of knowledge is its epistemological representation through subject-object relations. When a person acts as the subject of cognition of various objects in the world, no contradictions arise: cognition is carried out by a subject with subjectivity. The situation is different in the case when cognition is carried out by an artificial intelligence that does not possess subjectivity. Can a non-subjective artificial intelligence be a subject of cognition? Or maybe there is no problem here and everything boils down to terminological ambiguity, and subjectivity is not required from the subject of cognition? According to V. Decomb, it is necessary to recognize the multiple heterogeneity of the term "subject" due to "the inability to combine in a single concept all the uses of the properties "subject", "subjective", "self", "ego", "I", "self" and so on ..." ^{[6, p. 13]}.

In this article, we plan to consider a number of issues that the author considers important for philosophical cognition carried out when non-subjective artificial intelligence is included in the system of subject-object relations. These issues include: the possibility and necessity of non-subjective artificial intelligence to be a self-sufficient subject of cognition, the unification of a person with artificial intelligence as part of integral subjects of cognition, the possibility of an alternative interpretation of subject-object relations.

The problems of subject-object relations involving artificial intelligence

This problem can be formalized in the form of two areas of research: firstly, the use of artificial intelligence as one of the tools included in an integral cognitive system, the core of which is a person, and, secondly, the ability of artificial intelligence to be a self-sufficient subject of cognition.

A subject, as is known, can be not only a person endowed with subjectivity, but also a set of people (as biological individuals, social individuals or socio–cultural subjects - personalities) united by a community of needs into a group. In particular, organizations (economic, political, educational, religious, etc.) that unite people and their means of cognition and creation of the world can act as subjects. These means are accumulated material and intellectual values, including knowledge and various technical means. An important component of the latter are computing systems, in particular, machine learning systems and more complex artificial intelligence systems.

Cognition and creation of the world is currently impossible without the use of the entire range of information technologies designed to expand human capabilities (and in the future, to a greater or lesser extent replace humans) in solving problems related to processing large amounts of information. It can be stated that integral subjects of cognition, including humans and computing systems (including artificial intelligence in the form of machine learning systems), are already relevant and widespread. At the same time, the leading role in such systems is still reserved for a human (human operator).

Artificial intelligence, even without subjectivity (self), is able to formulate the specific subtasks required for its solution within the framework of a human task. Depending on the level of artificial intelligence, its ability to go beyond the initially set task varies significantly. There is no clear limit to the capabilities of artificial intelligence, including intelligence that does not possess subjectivity. The absence of subjectivity, however, imposes a limitation that is associated with the initialization of activity (including cognitive and creative) in order to meet the needs that a person endowed with subjectivity has, but non-subjective artificial intelligence does not.  

The ability of artificial intelligence to go beyond the initially set task is determined by the degree of its autonomy. The question of the degree of autonomy of artificial intelligence is a complex one, consisting of two main components.

The first component of the question of the degree of autonomy of artificial intelligence boils down to the division of autonomy into three levels. Let's call them the level of imperative programming, the level of declarative programming, and the level of creative artificial intelligence.

The first level corresponds to the minimum autonomy required to perform detailed step-by-step instructions (programs) generated by a human operator. The formation of additional subtasks, correction of the initial task in the process of execution are not expected. This degree of autonomy correlates with the imperative programming paradigm ^[7].

The second level of autonomy of artificial intelligence corresponds to its ability to form tools for achieving human goals. The specified tools can be formed at the software level or include the capabilities of hardware expansion of artificial intelligence in accordance with the needs arising within the framework of the task being performed. The second level of autonomy correlates with the declarative programming paradigm ^[7].

The third, the highest level of autonomy of artificial intelligence, corresponds to its ability to set intellectual tasks. For such a creative artificial intelligence, the problem of subjectivity becomes relevant. However, as noted above, artificial intelligence does not necessarily have subjectivity to fulfill its functional purpose.

The second component of the question of the degree of autonomy of artificial intelligence is to determine the required level of autonomy for solving various tasks. The main function of cognitive systems, of which artificial intelligence systems are a special case, is to replace humans in controlling machines that produce goods (material and spiritual values) to meet people's needs ^[8]. To solve the vast majority of machine control tasks, up to process and production control systems, automated and robotic digital productions, the capabilities of artificial intelligence systems at the level of imperative or declarative programming are sufficient. To solve some tasks related to processing large amounts of data (for example, in image or speech recognition, collection, classification and clustering of heterogeneous information, etc.), creative artificial intelligence is needed, which is usually based on artificial neural networks. However, even in this case, its subjectivity is not necessary.

Thus, at the first two levels of autonomy of artificial intelligence systems, they act as components of integral subjects of cognition, in which the leading role is reserved for the human operator. In the case of the implementation of creative artificial intelligence, both an integral subject of cognition, including a human operator and a subordinate non-subjective artificial intelligence, and a fully autonomous artificial intelligence are possible. In the latter case, artificial intelligence can remain non-subjective without loss of functionality. Tasks for such a non-subjective artificial intelligence are generated (but not necessarily formulated as a task) by the needs of a person endowed with subjectivity.

Why does the author so persistently and consistently promote the idea of rejecting the subjectivity of artificial intelligence? There are three main reasons. The first reason, which we have already mentioned several times above, is that, as research shows, there is no functional need for the subjectivity of artificial intelligence ^[3]. The second reason is the potential danger to humanity of artificial intelligence, which has subjectivity, which implies that it has needs (as well as, possibly, desires and emotions), which in many ways will not coincide with the needs of humanity. The third, no less important reason, is the need to preserve man's leading role in civilization: the role of initiator (actor) of all significant processes of cognition and creation of the world ^[8].

Cognition through artificial intelligence

One of the most common approaches to the definition of cognition is a definition that binds it to human intellectual activity: "Cognition is a specific type of spiritual and practically transformative human activity, consisting in obtaining, accumulating and transferring knowledge from generation to generation about both the properties, states, trends and patterns of development of the world around a person, and about the peculiarities of the existence of man himself in this world" ^[9].

The development of information technology, which opens up the possibility of creating artificial intelligence, makes this binding optional. As a result, an expanded representation of cognition is formed, carried out by a subject of any nature: both human and artificial intelligence. The following author's definition corresponds to the expanded view: cognition is the process of forming an expanded model of reality in consciousness. At the same time, an essential part of the expansion of the reality model is the result of creative activity. The proposed formulation of cognition correlates with the definition of consciousness within the framework of the information concept, according to which consciousness is an information environment in which an expanded model of reality is implemented ^[10]. A similar interpretation of consciousness is proposed within the framework of the concept of information synergetics ^[11].

The information concept of consciousness presupposes the following connection of consciousness, self-awareness and subjectivity. Under certain conditions (when consciousness is localized in the space of states of the carrier of consciousness), consciousness acquires the property of self-consciousness, a special case of which (in the case when the initiator of changes determining localization is the carrier of consciousness) is self-consciousness endowed with subjectivity ^[10].

The ability of the intellect to act as a subject of cognition is formed not when it acquires subjectivity, but at a significantly lower level – during the formation of self-awareness. The difference between self–awareness and subjectivity is a question that has not yet been definitively answered.

In our opinion, as already noted, subjectivity "grows" out of self-awareness, but is not identical to it. Subjectivity is a property of self–awareness endowed with the ability to be an actor of one's own changes and changes in the surrounding world. The initialization of the activity of self–consciousness endowed with subjectivity is the result of will ^[12] or desire, in turn, conditioned by needs.

The concept of self-awareness is somewhat more complex. According to the most general formal definition, self–awareness is the ability to separate oneself from others, "I" from "not-I". More detailed studies have shown ^[10] that the central attribute of self-awareness is the definition of its boundaries in the form of localization of the carrier of consciousness (human or artificial neural network implementing artificial intelligence) in the multidimensional space of states of the original real objects, as well as their reflections in consciousness in the form of information objects. Localization is provided by using feedback between the properties of real objects and their reflections in consciousness. Through them, consciousness "defines" the boundaries between the carrier of consciousness and the surrounding world. This is how consciousness becomes aware of itself (its carrier).

Self–awareness is a necessary property of the subject of cognition, without which it is impossible to separate it from the objects of cognition. At the same time, in a special case, the object of cognition can be the subject of cognition himself, who is aware of his own boundaries: "a subject is a being who has the opportunity to become an object in relation to himself" ^{[6, p. 102]}.

A cognitive system is a multi–level system that performs the functions of recognizing and remembering information, making decisions, storing, explaining, understanding and producing new knowledge ^[13], special cases of which are natural and artificial intelligence, does not necessarily have to have great capabilities in solving intellectual or computational problems. Its distinctive feature is that the cognitive system is a cognizing system (from Latin cognitio – knowledge, cognition), which means it must have self–awareness. And this is not such a difficult task. Insects have self-awareness, they are able to react to external influences, while distinguishing themselves from the environment. Self-awareness (in our accepted interpretation) is possessed by a simple artificial cognitive system, for example, a car parking system capable of identifying (separating one's car from others) and determining one's position in space relative to other cars and parking places.

Thus, it can be stated that in order to act as a subject of cognition, artificial intelligence does not require subjectivity or great abilities to solve intellectual problems, but it is necessary to have self-awareness.

It is obvious that the statement of the necessity for the subject of cognition to have self–awareness is a formal consequence of the definition of the concept of cognition in the logic of the opposition of subject and object. The validity of this approach is beyond doubt. However, is this the only possible approach? If we proceed from a different interpretation of cognition and, accordingly, subject-object relations (up to the recognition of the insolvency of such a representation), perhaps our conclusions about the requirements for the subject of cognition may change?

The interpretation of subject-object relations, which is based on the opposition of the subject and the object of knowledge, is called epistemological. The epistemological interpretation of subject-object relations is accepted within the framework of Cartesianism, classical and non-classical epistemology. The latter can also include post-non-classical epistemology ^[14].

Within the framework of Cartesian dualism (p. Descartes) postulates the presence of two substances – thinking and extended ^{[15, p. 128]}. "... classical epistemology proceeds from the idea of the eternal nature of the epistemological (i.e., cognizing) subject and the immutable laws and properties of cognition; ... non–classical - proceeds from the idea of a real living, historically and socially defined subject, armed and limited by its language, a subject capable not only of cognition, but also of delusion; The properties and laws of cognition are considered as historically changeable" ^{[16, p. 21]}. The approaches of classical epistemology (F. Bacon, I. Kant, etc.) were developed in the phenomenology of E. Husserl ^{[17, p. 79]}, the approaches of non–classical epistemology - in Marxist epistemology ^[18], which assumes additional consideration of the influence of the activity of the subject (person) on the objects of knowledge.

An alternative interpretation of subject-object relations is the ontological one. This interpretation, in particular, was put forward by Hegel in his works. Hegel considered the problem of the relationship between subject and object through the prism of alienation. The spirit (the subject of cognition), passing through a series of stages of objectification (alienation) and de-objectification (removal of alienation), realizes the object of cognition as its creation. As a result, there is no opposition between the subject and the object: as a result of the removal of alienation, the object merges ("absorbed") with the subject ^{[19, p. 15]}.

M. Heidegger proposed his own version of the ontological interpretation of subject-object relations. According to M. Heidegger, the subject "collects everything on himself as a foundation. This metaphysical meaning of the concept of the subject has no immediate underlined relation to the person, and even more so to the Self" ^{[20, p. 48]}. The subject acts as the basis on which the picture of the world is built, which M. Heidegger identifies with the world. As the picture of the world develops, the subject is formed, which "becomes a representative of what exists in the sense of the objective" ^{[20, p. 50]}.

The ontological interpretation of subject-object relations identifies being and consciousness. At the same time, a necessary and sufficient property of consciousness for removing alienation (according to Hegel) or building a picture of the world on the subject as the basis (according to Heidegger) is its ability to cognition. However, if the subject and the object are not separated (unlike the variant of the epistemological interpretation of subject-object relations), then how does cognition occur and does the subject of cognition have self-awareness? A consistent explanation for the realization of self-awareness within the framework of the ontological interpretation of subject-object relations is the idea that consciousness unites everything and that there is only one subject of cognition (a person or a Spirit) who knows himself. That is, the ontological interpretation of subject-object relations leads us to idealism: absolute – in Hegel, subjective (in the form of existentialism) – in Heidegger. The idealistic understanding of the world corresponds to the philosophy of irrationalism, within which, as we have already noted, we cannot get a reliable answer to the question of the need for subjectivity to solve intellectual problems.

If we set the task of using non-subjective artificial intelligence for cognition and creation of the world, we are forced to choose an epistemological interpretation of subject-object relations: firstly, based on the reasonableness of ideas about a deterministic world; secondly, due to the impossibility of getting answers to the questions we are interested in within the framework of an ontological interpretation.

Conclusions

Let's summarize the research conducted in the article:

1. The development of information technology has now created the necessary conditions for the creation of artificial intelligence systems capable of replacing humans in solving a wide range of intellectual tasks. A significant feature of artificial intelligence is the presence or absence of subjectivity. In the latter case, there is probably no decrease in its functional capabilities, but there is a limitation of its ability to independently initialize cognitive or creative activity.

2. Depending on the degree of autonomy of a non-subjective artificial intelligence, its position in the system of subject-object relations may consist in participation in an integrated subject of cognition, in which the leading role is reserved for the human operator, or in the completely autonomous implementation of cognitive or creative activity to meet (formalized or non-formalized) human needs.

3. In order to carry out cognitive and creative activities, artificial intelligence, like any other cognitive system, does not require subjectivity, but self-awareness is necessary, which is based on the ability to separate oneself (the subject of cognition) from the objects of cognition.

4. The epistemological interpretation of subject-object relations is not the only possible one. There is also an ontological interpretation in which the subject and the object of knowledge are not opposed. This interpretation, however, inevitably leads to idealism and irrationalism, which makes it impossible to determine the need for subjectivity and the potential capabilities of artificial intelligence on the basis of this interpretation.