THE BRAIN AS A CIRCUIT DEVICE

Petriychuk N.D.

Moscow, Russia
e-mail: sergtoolserg@yandex.ru

Any theory, following scientific methodology, must be supported by evidence. This minimal criterion presents a challenging verification challenge for the study of consciousness, as we do not know whether a physical system can be confirmed through external observation to be aware of what it perceives internally—an unsolved problem known as the "hard problem of consciousness." To arrive at a theory of consciousness, the hard problem has prompted the development of phenomenological approaches that accept assumptions about the properties consciousness exhibits based on personal experience and, from these assumptions, infer the physical processes that cause these properties to manifest. Although consciousness, as an immaterial phenomenon, lacks any properties, its presence can be manifested by certain properties in the mechanisms that implement it. A prominent theory employing this approach is integrated information theory (IIT), which proposes that our subjective experience is a unified whole, subsequently necessitating physical feedback as a condition for the emergence of consciousness.

Every sufficiently detailed theory of consciousness, including IIT, must be isomorphism-invariant to avoid the existence of isomorphic "philosophical zombies" and the epistemological problems they pose. It should be clarified that "philosophical zombies" andwhether "«аThe "conceivability argument" is a thought experiment in the philosophy of mind designed to challenge the physicalist approach. Physicalists, of course, reject this argument, but that doesn't mean it's generally meaningless. The debate surrounding this argument offers valuable insights into other issues—modal logic, the concepts of the possible, and the conceivable. A philosopher shouldn't unconditionally accept the notion that consciousness is secondary to the brain, nor should they unconditionally accept the opposite argument, because the philosopher's task is to develop a variety of approaches beyond the most obvious, since the most obvious isn't always better than something else. Philosophers have a concept of immaterial form, and consciousness can be understood as the immaterial form of brain processes when these processes are brought under subjective attention, thus informing the essence of what is being understood.

The integrated information theory, proposed by neuroscientist J. Tononi, explains consciousness as integrated information, i.e., an informational state produced by the entire system from the interacting set of its constituent parts. Many philosophers are attracted to this approach because it offers a way to explain consciousness in a concrete, systematic manner. IIT implies that consciousness is epiphenomenal and is generated by the subjective meaning of integrated information, rather than being a measurable property [9].

The term "hard problem" was coined in 1995 by Australian philosopher David Chalmers [2]. It suggests that consciousness cannot be explained in functional terms. Thus, to solve the easy problems that explain cognitive abilities and functions, one only needs to specify the mechanism that can perform that function. In contrast, the hard problem is unique in that it is not a problem of functional performance; its "puzzle" persists even after this performance is explained—remarkably, the physical processes of creating alternative automatisms for new conditions in the brain are accompanied by consciousness. This could be called an internal subjective "movie" or, better yet, a sense of self.

Integrated information theory takes the existence of consciousness as a given. By examining the subjective manifestations that characterize consciousness, it suggests that they can tell us what properties and functions a being's brain must possess if we are to confirm its intelligence. Therefore, G. Tononi prioritized defining the properties inherent in conscious mental states and then assumed that everything that embodies them must share these corresponding characteristics. He called the former axioms, the latter postulates.

The axioms describe how conscious states exist, have a specific structure, contain integrated information, and have exceptions, such as the speed with which they arise—this must be true of any physical system considered conscious. Given the theory's name, this raises the question: what does it mean for a system to have integrated information? According to Tononi, information (e.g., about visual experience) is integrated because it cannot be separated from the whole. Take, for example, a blue book. We cannot see the book and the color separately—we only see the whole picture—the blue book. Separating these two components (the color and the object) is impossible. Tononi argues that information in the corresponding brain structure is integrated in the same way, meaning that its parts are causally related; if you separate them, the information disappears (as in the book example).

Integrated information, each of its components, can in principle be measured. Integrated information theorists propose measuring it using the PFI function [12]. Therefore, if the theory is correct and intelligence is a system of integrated awareness for adaptation purposes, we can determine whether a machine is intelligent in principle. To do this, we need to look at its "positronic brain" and calculate the PFI. If the PFI value is as high as that of a human, then the machine is intelligent. However, in the experience of actually creating adaptive systems, this approach to verifying intelligence proves to be unacceptably simplistic. Not to mention that there is currently no way to measure the PFI of a human brain, much less a machine.

It seems that the main difference between integrated information theory and other theories that attempt to explain the mind lies in their approach to explaining whether a being has a mind or is simply self-aware. Most theories primarily attempt to figure out how brain The mind produces it, and integrated information theory, as stated above, first attempts to explain the properties of the mind. However, there are no terminologically correct definitions of mind and consciousness, which is generally characteristic of theories of philosophical quality.

By studying the physiology of the brain during specific behaviors (e.g., sleep), scientists have discovered anatomical locations of activity corresponding to characteristic manifestations of the conscious state—"neural correlates of consciousness" (NCCs)—that indicate the location in the brain of the mechanisms underlying conscious experience. On the other hand, NCCs cannot be used to explain why we are conscious or to predict whether another system exhibiting similar properties to NCCs is conscious. In fact, NCCs can only tell us about physiological processes that correlate with the putative functional consequences of subjective self-perception and, in principle, may not adequately measure subjective experience. In other words, we can objectively measure behavior that we assume accurately reflects consciousness, but currently lack the scientific tools to test our assumptions. As a result, we struggle to discern whether a system is truly conscious or simply making movements, producing external signs, or even actively reporting an internal experience that is not actually present.

Phenomenologically, consciousness is often associated with the concept of "downward causation," whereby "higher-level" mental states exert causal control over lower-level realizations. According to this view, the "additional information" provided by the processes that generate the subjective sense of self is considered functionally significant, as it influences how some states transition to others. It is important to acknowledge that there is an alternative view, in which differences related to consciousness are defined not in terms of abstract computations, but in terms of concrete circuit implementations, as accepted by IIT. However, this does not address the question of whether isomorphic systems ultimately experience phenomenological differences, as there is no other way to test this assumption, and it remains merely to postulate it.

For a complex system like the human brain, there presumably exist many possible circuit architectures that define an equivalence class capable of performing the same functions with the same input, differing only in how the states are represented internally (for example, how neurons are interconnected). This could, for example, explain why the human brain normally possesses a subjective sense of self, despite differences in the specific features of the neuronal network structure, and why the internal representations that evolved were selected in the first place. But this would not lead to an understanding of why the effect of subjective self-awareness arose in the first place. Just as it would be impossible to understand why and how the effect of a program running on a computer monitor arose solely from the observable image on the monitor.

We must proceed not from the end of the chain of known facts, not from consciousness, trying to untangle the tangle of causes, but from the functionality that led to the emergence of an "integrated information" system—to the set of recognized, diverse features of perception and possible actions under given conditions, their comparison, and the identification of the most important—as a consequence of the advantages of a given level of adaptability. It is precisely this understanding that makes practical results possible in modeling the individual adaptability system. This opens up new ways to solve the "complex problem" of consciousness.

Based on the theoretical model of the Volitional Adaptability Model of the Psyche [5], a software implementation of a system of principles of individual adaptability was implemented. Modeling was accomplished by sequentially adding new adaptive enhancements to existing ones, repeating the stages of evolutionary complexity in organisms from the simplest insects to species possessing a psyche. This development included the following levels of qualitative improvement in adaptation effectiveness:

• the level of unconditioned reflexes of homeostasis based on the deviation of vital parameters from the norm with the formation of basic contexts of perception-response - behavioral contexts;
• the level of unconditioned reflexes to the simplest stimuli;
• the level of unconditioned reflexes based on the previous ones – instincts;
• the level of conditioned reflexes based on previous ones with the fixation of new trigger stimuli;
• the level of identification of significant novelty in the features of perception depending on the current activity of basic contexts – orienting reflex;
• the level of formation of subjective contexts in the form of a «tree of understanding», reflecting current conditions, including influences, with a connection to the most habitual reactions, which are created as an alternative to reflexes and other habitual reactions in conditions of significant novelty, when old reactions turn out to be inappropriate;
• organization of memorization of episodes of drawing attention to significant novelty in memory - episodic memory, which has prognostic properties;
• organization of a system of target motivation depending on the current state;
• organization of a system for storing the experience of attempts at behavioral responses in given conditions with an assessment of their effectiveness;
• functions of supporting the connection of conventional symbols of perception with the existing meanings of images of perception and action - vocabulary for organizing the possibility of interaction with another organism;
• the level of formation of the current integrated environment of awareness, consisting of various types of recognizable data, similar in idea to that in the interpretation of G. Tononi [7], which includes an active branch of the «tree of understanding», a comparison of individual recognized elements of perception into a generalized meaning (subjectively assessed significance), episodic memory, target motivation, experience of implementing behavioral options in the past, forecasts of possible results based on episodic memory;
• functions of choosing the most preferable course of action in current conditions of significant novelty;
• functions of choosing an option of action in the context of search behavior to obtain results in current conditions of significant novelty and the formation of a habitual response option;
• functions of creating a dominant of an unresolved problem in case of difficulties in choosing the preferred behavior.

All of these functional mechanisms and their interactions are described in detail in the report [4], clearly revealing the adaptive functionality of each. The model operates in real time and is not an emulation. As the project developed, it became clear that a natural neural network is, first and foremost, a sequential layering of local adaptive mechanisms that orchestrate a response (external and internal) depending on the conditions.

Subsequent levels of such formations gain the ability to voluntarily influence previous ones (functional feedback), the simplest example of which is the retention of a stimulus by an orienting reflex through the cycles described by A. Ivanitsky [3]. There is a huge number of such sequentially complementary links, and all of them are "hereditarily" predetermined.

During development, numerous opportunities arose to "improve" and modify the organization of adaptive systems, starting from the very foundations. Evidence from studies of natural implementation, using a spectrum of evolutionarily complex living beings implemented in circuitry, confirms that the latter will have advantages over biological beings in every way. The idea emerged to base the understanding of the psyche not on the vague concept of "consciousness," but on the phenomenon of ongoing self-awareness, which is present in all creatures possessing a system for processing the abstracting "orienting reflex," i.e., possessing a hippocampus.

The Bot diagram (the name of the software model) presents a sequence of human development through all the necessary stages. Self-awareness is a well-formalized concept [1], expressed by a specific structure that integrates data from the current state:

• «The »tree of understanding” identifies conditions that correspond to the moment of influence on the Bot of internal and external factors, as well as the results of one’s own arbitrariness;
• generalization of individual features of perception into a common meaning (subjectively assessed significance;
• episodes of memory of similar situations with a forecast of possible consequences;
• a range of possible actions that are customary for a given specific set of conditions;
• target motivation corresponding to the given conditions;
• a set of existing experience in developing habitual actions in the past with successful results, which is used in choosing alternative behavior options when there is an important novelty that predicts the failure of a habitual reaction in given conditions;
• active dominants of unresolved problems (if any);
• a mechanism for developing goals in given conditions (if there is a need to do something).

All these components together create a «picture» of the current self-perception, into which new details break through as the «orienting reflex» is triggered, including new results of data processing (ideas) outside this integral area of awareness.

The key finding of this study is that there is nothing fundamentally fundamental about the organization of the brain's adaptive mechanisms, requiring entirely new, complementary concepts and ideas. The very possibility of successfully implementing a system of adaptive functions, which is generally recognized as inherent to the level of subjective information processing, demonstrates that the brain is organized precisely as a circuit-based device, that is, it represents cause-and-effect mechanisms, and nothing else. This confirms the following assertion: the circuit-based approach not only justifies itself, but is an effective approach to understanding the mechanisms of higher-level adaptivity [6].

Thus, for the first time, a system of a large number of known and new psychophysiological mechanisms, holistically interacting within the framework of individual adaptability, was implemented in a schematic manner. Previously known phenomena received a concrete embodiment and clear functionality.

It has been shown that an individual adaptive system can be unique only if homeostasis exists, optimizing vital parameters, simply because otherwise there is no criterion for adaptability: there are no parameters that need to be or can be maintained within the norm (the basis of any adaptive system). It is the circuit-based approach that adequately describes the brain's adaptive system [8] – the hierarchy of principles of individual adaptability can be organized independently of the method by which its mechanisms are implemented.

The work also demonstrated the fundamental validity of the philosophical assumptions formulated by J. Tononi, who asserts that the phenomenon of awareness is based on an integrated environment in which information from many different components of the current state converges. The principle of hierarchical behavioral contexts offers enormous potential for reflecting any conditions with any response, something that game-based artificial intelligence lacks.

Using a circuit design approach, it is possible to create different systems of living beings (having their own homeostat).

Here we approach an important question. If we assume that adaptability at the level of consciousness arises from a homeostatic mechanism, then how does homeostasis itself arise? The answer to this question can lead to an abstraction that goes beyond psychological and physiological "manifestations"—free energy minimization. For self-organizing systems to exist, they must have mechanisms to counteract the thermodynamic tendency of local entropy (which can be interpreted as the amount of free energy of a local system). That is, self-organizing systems can exist over long periods of time only by occupying "preferred" states—this is a fundamental prerequisite for life and, indeed, for any self-organization. Here, we need not concern ourselves with how life arises. However, the justification for a mechanism subjectively taking the form of consciousness is a necessary prerequisite for life at a certain level of self-organization, since it is generally assumed that everything conscious is alive, although not all living beings are conscious.

For a system to resist entropy, certain conditions must be met. These conditions allow self-organizing systems to exist and persist over time and arise naturally within any ergodic system. A randomly chosen dynamic system has a "Markov blanket" that sets the conditions and is a probabilistic-statistical construct that depends on what influences what and what does not influence what [11]. It is important to realize that sensory and active capacities are properties of the blanket, not of the states with which they interact, which implies that a system isolated by a Markov blanket can "know" states outside the system only indirectly (which is essentially what Gödel's theorem says) [10]. In other words, external states can be "inferred" by the system only on the basis of "sensory impressions" on the Markov blanket. In fact, in order to resist dissipative forces, it is important that external states are sufficiently adequately recognized by the system. This means that the system must incorporate a model of the world within its habitat, which then becomes the basis upon which it operates. Such systems are the best possible implementation for ensuring individual adaptability, which must, first and foremost, take into account local environmental conditions. This is why they have been distinguished by natural selection.

The inferences the model (Bot) generates for the system about external conditions (inferences formed based on the sensory consequences of its actions) take the form of predictions, and these predictions must be constantly tested and revised. Perception and action entail constant processes of hypothesis testing, through which the system experimentally updates its model (its "beliefs") over time—for this, the system has genetically ingrained foundations in the form of an exploratory behavioral style and, at the psychological level, the emotion of curiosity.

Whether homeostasis functionally minimizes the free energy of a system, as a special case of the more general phenomenon of self-organization, is a debatable question, because at the level of arbitrariness, behavior is possible that is subjectively assessed as extremely important, but in doing so it loses not only energy, but also life itself.

Currently, the Bot model demonstrates adaptability comparable to that of a 3-5-year-old child, and in some respects (such as proactively selecting the best option), even higher. Further progress requires significant periods of development of existing mechanisms (accumulation of a vocabulary of meaningful words, accumulation of episodic memory for predictions, and accumulation of experience in response). After this, the project must implement mechanisms of creativity and experience in voluntary action (in humans, the process of developing this level of experience takes up to 25-35 years). Therefore, rapid methods of information delivery are insufficient – interaction with the operator occurs in real time, and both parties must engage in lively, adaptive communication.

The purpose of developing this model is to demonstrate the possibilities and full achievability of a circuit implementation of a system of adaptive mechanisms (individually long known to neurobiologists), a system in which the emergence of each new way of being informed about what is happening provides new qualitative potential for adaptation mechanisms.

The bot evolved not by adapting to an existing model, but by gradually accumulating functional capabilities at each new level, developing at each stage useful (for adaptability purposes) potentials from the combined capabilities of the previous stage. An additional goal was to compare the principles of adaptability—natural and software implementation.

In conclusion, it should be noted that this article does not aim to solve the "hard problem" of consciousness. However, the modeling conducted provides a perfectly clear basis for understanding the essence of this phenomenon. The modeling demonstrated that at the level of creating new automatisms in new conditions—as an alternative to the familiar—an integrated information environment is necessary, abstracted from the immediacy of perception and revealing a set of parameters that determine the significance of the current state for the subject. This environment is capable of interpreting primary perception according to the emerging understanding of the meaning of what is happening, with arbitrary feedback, so that the most incredible perceptual illusions can arise.

Literature

1. Basic Self-Awareness [Electronic Resource] // Access Mode: fornit.ru/160 (Accessed: 23/12/2021)
2. Demin T.S. David Chalmers's Conceivability Argument // Philosophical Journal. 2020. No. 3 [Electronic resource] // Access mode: https://cyberleninka.ru/article/n/argument-predstavimosti-devida-chalmersa (date of access: 23/12/2021)021

3. Ivanitsky A.M. «"Reading the Brain": Achievements, Prospects, and Ethical Issues. Pavlov Journal of Higher Nervous Activity. 2012. Vol. 62. No. 2. Pp. 133-142.
4. Brain and Bot [Electronic resource] // Access mode: fornit.ru/49721 (date accessed: 2312.2021)
5. Model of Volitional Adaptability of the Psyche [Electronic resource] // Access mode: fornit.ru/7431 (date accessed: 23.12.2021)
6. Circuit design of adaptive neural networks [Electronic resource] // Access mode: fornit.ru/41930 (date accessed: 12/23/2021)
7. Tononi J. Integrated Information Theory [Electronic resource] // Access mode: fornit.ru/7589 (accessed: 23.12.2021)
8. What is the I-schematic approach [Electronic resource] // Access mode: fornit.ru/40830 (date accessed: 2312.2021)
9. Yashin A.S., Dubynin I.A. G. Tononi's Theory of Consciousness: Analysis and Criticism [Electronic resource] // Access mode: https://elibrary.ru/item.asp?id=46567990 (date accessed: 23.12.2021)
10. Batzoglou S. Gödel's Incompleteness Theorem [Electronic resource] // Access mode: https://arxiv.org/pdf/2112.06641.pdf (Accessed: 23/12/2021)
11. Kirchhoff M., Parr T., Palacios E., Ariston K., Kiverstein J., 2018 Markov Blankets of Life: Autonomy, Active Inference and the Principle of Free Energy JR Soc. Interface. 15 20170792 20170792 [Electronic resource] // Access mode: http://doi.org/10.1098/rsif.2017.0792 (date accessed: 23.12.2021)
12. URL: https://docs.microsoft.com/en-us/dotnet/machine-learning/how-to-guides/explain-machine-learning-model-permutation-feature-importance-ml-net (Accessed: 23/12/2021)

References

1. Basic self-healing [Electronic resource] // Access mode: fornit.ru/160 (data obrashhenija: 2312.2021)
2. Demin TS. The Argument of Djevid Chalmers's Representability // Filosofskij zhurnal. 2020. No. 3 [Electronic resource] // Access mode: https://cyberleninka.ru/article/n/argument-predstavimosti-devida-chalmersa (data obrashhenija: 2312.2021)021
3. Ivanickij AM. «Brain Reading: Achievements, Prospects, and Jet Problems. I. Pavlov Journal of Higher Nervous Activity. 2012. Vol. 62. No. 2. P. 133-142.
4. Brain and Bot [Electronic resource] // Access mode: fornit.ru/49721 (data obrashhenija: 2312.2021)
5. Model of Volitional Adaptability of the Psyche [Electronic resource] // Access mode: fornit.ru/7431 (data address: 2312.2021)
6. Shemotechnics of adaptive neurons [Electronic resource] // Access mode: fornit.ru/41930 (data obrashhenija: 23.12.2021)
7. Tononi J. Theory of Integrated Information [Electronic Resource] // Access mode: fornit.ru/7589 (data obrashhenija: 2312.2021)
8. What is the Ja-shemotehnicheskij podhod [Electronic resource] // Access mode: fornit.ru/40830 (data obrashhenija: 2312.2021)
9. Jashin AS, Dubynin IA. J. Tononi's Theory of Consciousness: Analysis and Criticism [Electronic resource] // Access mode: https://elibrary.ru/item.asp?id=46567990 (data obrashhenija: 23.12.2021)
10. Batzoglou S. Gödel's Incompleteness Theorem [Jelektronnyj resurs] // Access mode: https://arxiv.org/pdf/2112.06641.pdf (data obrashhenija: 2312.2021)
11. Kirchhoff M., Parr T., Palacios E., Ariston K., Kiverstein J., 2018 Markov Blankets of Life: Autonomy, Active Inference, and the Principle of Free Energy JR Soc. Interface. 15 20170792 20170792 [Jelektronnyj resurs] // Access mode: http://doi.org/10.1098/rsif.2017.0792 (data obrashhenija: 23.12.2021)
12. URL: https://docs.microsoft.com/en-us/dotnet/machine-learning/how-to-guides/explain-machine-learning-model-permutation-feature-importance-ml-net (data obrashhenija: 2312.2021)