Hardware and software of the brain

The sentence represents some action denoted by its verb with several objects denoted by noun phrases as participants. These objects are linked by relations denoted by prepositions. In principle, this method of constructing a complex object with prepositions may be used without any verb at all. 'A cup on the table'. 'A box under the chair'. 'Instruments in the box'. This renders a quite definite scene. A normal sentence may be regarded as a bag (container represented by the verb) full of such relations.

Homogeneous members of the sentence present a major difference from programming languages. In the latter, some variables may be declared as lists. In natural language, most of members of the sentence are lists by default. This is normal because, say, each noun usually has several attributes. If just one of them is explicitly specified, this is a list with just one element.

In mathematics, variables are placeholders used for 2 different purposes. In equations, they denote unknown values and work for output. In functions, they represent input parameters. In programming, such parameters may be passed by value or by reference. In the latter case, the placeholder works as a link in hypertext. Similar methods may be discovered in natural language.

Pronouns 'it' or 'this' are usually function as a short-distance reference. This case shows clear difference in overall approach of programming and natural language. The last has no strict rules of reference. Oftentimes it is ambiguous and confusing. Meanwhile it may be very useful. For example, a whole paragraph may be devoted to the description of some object. Later, this object may be used via 'this'.

Another method employs generic names. In Object Oriented Programming, the term object denotes a definite data structure while its generalized description is called a class. Before using functionality of a class, you must create some object. Natural language is less restrictive. 'The table' refers to the whole class and you can use it in a sentence. You can use 'this table' so as to select one particular object from this class. You can also use filters to create a subclass 'tables which were purchased for this office'.

Modern languages can generate very long, complicated sentences, but in most cases they may be reduced to a list of simple sentences. The aforementioned variables are used to link them. 'Tables which were purchased for this office arrive to the back door of the building.' -> '20 tables were purchased for this office. The office building has a back door. The tables arrive to the door.' Here the definite articles turn generic names into particular objects. An understanding machine may create a separate internal variable for each more or less significant object of the text. Later use this variable in the semantic representation or update the description of the object.

Copula verbs represent the unique static action, that is a state when nothing changes. Its main verb – 'is'. In principle its syntax is not very different from other verbs. 'He is a sportsman.' Here 'a sportsman' may be parsed as a direct object, but semantically it is very different. This represents conceptual hierarchy, that is what is called class derivation in OOP. Yet another semantics requires a syntax unknown for normal verbs. 'The apple is yellow.' Here the predicate phrase consists only of the adjective. This is equivalent to the sentence without any subject: 'Yellow apple.' with stress on the noun rather than the adjective. It is used so as to set attributes of an object.

Other copula verbs are: become, grow, get, turn, appear, look, seem, etc. The first in this list denote the transitional action of switching the state. The last adds semantics of perception. 'Is' denotes the absolute truth. The other verbs acknowledge that this is only as we perceive it. The last even expresses some doubt. Such samples only underline that our brain is an analog computer which allows infinite variations.

Formalized programming vs. free human communication

In the first case everything should be explicitly defined. Instead, humans widely use defaults and implied meaning. When one man formulates a task for another, he usually doesn't mention many obvious things. Moreover, the task is usually outlined in general. It is up to the employee to add necessary details in the process of execution depending upon the current situation.

Summary

Human language is a product of evolution. It can't be perfect already because it is permanently in development. Various features are being added by different people in different circumstances. It is essential to understand that live language is no more than a pile of raw material. In addition, it works on an analog computer – the live brain. If we want a workable discrete system for standard computers, we need a consistent theory of the language first. Then – its implementation. This theory can't embrace all the features of unlimited human language in principle. Because this language is internally contradictory. Its implementation would be unworkable. Thus the term "natural language" – a workable extract which may be specified explicitly in full. Let's do it.

Natural language is used to describe knowledge. It is arranged in hierarchy: library – domains – books – chapters – paragraphs – sentences – clauses – phrases – words. When we read a text, it is translated into some internal semantical representation, that is meaning. The process is called understanding. The structure of this representation is a key problem. If we know it, it is always possible to arrange translation. The main application of this representation is problem solving. Question answering and search is just a subtask.

Thorough analysis of natural syntax discovers just a few basic semantical elements. Each sentence represents some action denoted by its verb. The whole text answers the question: "What happens around?" The action may involve several objects, that is static elements. Essentially, each action may be represented by some object as a function of time. That's the difference. In the nervous system both are translated into static or dynamic neural images. In the sentence, objects are usually represented by noun phrases. The first of them is called the subject. It is used to underline this object and draw attention to it. Both actions and objects have attributes denoted by adverbs and adjectives respectively. Also these elements may be linked by relations. Prepositions represent relations between objects. Conjunctions – between actions. That's all. The whole of human thinking, problem solving, and complicated decision making is made of these elements.

Let's look what happens when we read a text sentence by sentence. First of all, the content may be subdivided into semantically different pieces. These categories are also semantical elements, but they belong to higher levels of the hierarchy. The exact composition may be different for different genres, but the principle is the same everywhere. I will consider various scientific textbooks as the most meaningful literature. For other genres, you may try and produce a similar listing yourself. Only keep in mind that some books are not intended to be meaningful at all. The author may stimulate filters of your mind which have the task not to let things in. The only goal is your entertainment.

The most often categories of content are: definitions of new concepts, presentation of facts such as a value of some world constant, composition of a complicated object (anatomy), listing of events in some process (history), making a statement (theorem), formulating rules which link events (if – then), etc. All these semantical elements contained in some pieces of text are intended for retention in your long-term memory. The procedure is like installing a program or database on the hard drive of a computer. Albeit adding information into the knowledge base is more complicated. At least, it should be checked for consistency. The new data should not contradict what already exists. What to do if such contradiction does emerge? You may either reject the new portion, or update your previous knowledge.

The most interesting case is when acquisition of new knowledge happens as a side effect. For example, 'Yesterday, he met his friend Nick who lives in the nearby town.' Accepting this sentence, the program should make a current record about the meeting plus make a new entry. It will create the new object named Nick and add the property that Nick is 'his' friend.

This example also demonstrates another function. The pronoun 'he' is a placeholder. It stands for some man. An understanding program should replace it with the corresponding name. This approach shows that the resulting semantical representation will be even larger than the initial text while intuitive assessment is opposite. What's the matter?

First of all, we retain only the upper levels of the semantical pyramid. This is obvious if you compare an abstract with the full text of some article. The summary tells you only what to expect there. For details, browse the article itself. Next, the text itself also contains supplementary pieces which are not intended for retention. For example, a theorem may be a valuable instrument for everyday use while its proof is needed only if you want to check whether it is correct. The amount of such content sharply increases if you recall that most of our knowledge is probabilistic. An important fact may be short but doubtful. Accordingly, it will be accompanied by lengthy commentaries with the only function – convincing the reader to accept that fact. Further, an important piece of knowledge is an instrument and the text may contain hints on when and how to use it better. These passages create a sort of index in your mind. They link the instrument with the situation where it may be helpful.

Modern languages allow very complicated constructs. One compound sentence may contain several subordinate clauses and occupy the whole paragraph. Looks like, our mind dissects them into the simplest forms of the nucleus language. It creates separate representation for encountered objects and processes (actions), replaces placeholders (variables) and lengthy references with concrete names. Also, it links these objects and actions into a semantical network using relations. This job is comparable with what is performed by a compiler when it processes the text of a computer program.

Consciousness

This is a long-lasting area of research attracting specialists from philosophy, theology, psychology, neuroscience. In recent decades computer scientists joined. When it comes to reproduce human ability in a machine, Artificial Intelligence is usually recalled. Machine Consciousness is yet another key feature.

Definition

In [1] it is perfectly demonstrated that prolonged debates leave major questions unresolved. This includes definitions of the main concepts. The situation is very characteristic for any science on initial stages of development. People like to introduce some term, even put it into practice, then discuss what it means. How is it possible? Computer programming strictly prohibits such practices. First define a variable, assign some value, only then you can use it. Fuzzy logic employed by neurocomputing in live humans is different. The initial definitions, even the move to recognize this concept is sub-conscious. Neural processes which underpin it are image-based, nonverbal. Accordingly, the first cases of usage may be highly erroneous, but with practice quality gradually improves. Discussion helps to verbalize the matters and exchange ideas.

The state of the art is that on one hand neuroscience accumulated a vast corpus of data about the structure and functions of the nervous system. On the other, various methods of data processing offer ready solutions. The computational approach to consciousness is based on the statement that the brain is a live automatic control system. Such devices are perfectly explored mathematically and this science guarantees a complete representation. All the possible solutions are known. We need only to choose which one is implemented in our own head. Optimal definitions may be formulated taking into account previous attempts and keeping in mind future use.

Different people use this term in a different sense. In the most general case, it is just all the higher nervous functions. A more narrow meaning is that consciousness is just upper levels of the perception hierarchy related to abstract thinking and understanding. Where to draw the lower boundary – decide yourself. One solution – just above the secondary sensory fields. Consciousness begins where sensory modality ends up.

The most narrow definition comes from neurocomputing.

Consciousness is self-control.

The basic principle of life is regulation or stabilization, self-support. The brain maintains the internal environment of the body and generates behavior. Only neurons are live as well. The brain must also maintain its integrity. It monitors own operation and corrects it when necessary. That's consciousness.

Macro regulator

Homeostasis is a founding principle of living nature. It is implemented by different means such as biochemistry. The nervous system adds to this toolbox. A classical regulator is described by a standard scheme.

Fig. 16. The standard regulator.

A sensor reads current values of the regulated parameter. The comparator outputs its difference from the normal value. Then, this difference produces a regulatory reaction. A similar scheme may be implemented for the whole organism, only instead of scalar values, sets of parameters arranged in 1D vectors or 2D matrices (images) are used. For example, you may maintain order in your living room or in the kitchen.

In this case, you will have an image of the norm, knowledge as the image of current reality, emotion produced by comparator, and some motor image of the regulatory output. This principle may be especially efficiently implemented in hardware using 2D neural nets.

The macro regulator is especially efficient to implement elements of consciousness on the hardware level. Brain operation is sophisticated. It is difficult to represent this by a single scalar parameter, even by a set of them, that is by a vector. The matrix representation is much more suitable.

Principles of neuroprogramming

Ivan Pavlov spent 20 years searching for physiological foundations of psychology and could not discover anything substantially different from reflex. Neural processes on higher levels of the hierarchy are different only by the fact that both the stimulus and reaction remain inside the skull. A reflex is replaced by an association between images. Programmatically, an association is a rule so we have a powerful rule-based machine. A part of what we know as software is implemented materially by hardware. White matter of the brain represents links between various cortical fields and subcortical nuclei. Each such link can store many rules and these pairs operate simultaneously in parallel processing. This provides sufficient foundation for sophisticated software. An internal knowledge base is a picture of the world used by this software.

Reason

Newborn humans are rather helpless. Most of our abilities were learned, that is programmed. Like in computers, human software is a complicated hierarchy: low-level routines, system programs, and high-level applications such as professional skills. Reason is associated with the second group. Obviously, it may vary among individuals and cultures, but there is something in common. Let's consider how to implement the reasonable execution of algorithmic applications. Another example of system software is a program for automatic problem solving. You may also add what you prefer.

Real-world execution

Our computers perform algorithms – predefined sequences. If something went wrong or real conditions don't fit what is required, the machine doesn't pay attention and makes an error. Humans are more reasonable. Accordingly, the procedure is more complicated.

Before each action we automatically check whether the current conditions are in accord with the prescribed. Also we project the consequences because the programmer could not foresee all the variants possible. This forecasting uses knowledge from the internal picture of the world.

After the action, we evaluate the real results and check whether they are acceptable. Otherwise, we must backtrack and redo the job.

Sometimes an obstacle emerges and the whole remaining tail of the algorithm becomes inadequate. Then, it must be reprogrammed in real time. This also uses the internal knowledge base.

Commentaries

As other major terms, the words from this cluster are de-facto used with a different meaning. One popular dualism is information versus activation. In the most primitive sense, consciousness is just non-sleeping. More subtle gradations are possible. We may be awake, but act automatically because upper levels of control are down. When they are on, consciousness as information processing begins.

Detailed schemes may be rather complicated. Let's consider, for example, decision making. There are 2 parallel channels: emotional and rational. In the first case, I do it just because I like this variant. In the second, the decision is guided by rules. Both cases may proceed consciously or subconsciously when we can't report why this decision was made. The specific mysterious feeling of the self emerges when the associative neocortex receives input from decision-making structures. We look at our own inner proceedings. Also, we may have a special area in the brain, the activation of which marks the state of wakefulness. When it is down, many (but not all) brain functions are switched off. Awakening is like pressing the Power button in a computer. The brain must know its current state. The feeling of the self serves this purpose.

Another example shows the usefulness of conscious self-control from the computational point. When we perform some activity, it is asynchronous. An action will be performed until the expected result is achieved. If this doesn't happen, the system will hang. Consciousness prevents it. As soon as we see that the action drags too long, a program of correction is turned on.

Religion

The common meaning of the English word conscience is emotional evaluation of own actions and is closely linked to religion. Also self-control but of a higher level. Furthermore, modern religions require the presence of a book. A written version transfers knowledge via the conscious channel. A new believer can acquire the basic facts and rules of behavior one by one. Indeed, religion is a network version of a knowledge base. On one hand, it defines individual behavior and the principles of interaction. On the other, one meaning of the term God is the soul of a nation. Religion defines a new living organism of the social level.

Cybernetics of consciousness

Full-scale consciousness emerges in advanced systems of multiparametrical regulation. When you have just a few biological needs, can manage them on the principle of domination (priority). Adding secondary needs already requires a vector norm. Advancing further to self-linked maintenance of brain workability shifts to a full-scale analog image of the self.

Human consciousness is the third regulatory loop of the brain. The first, most ancient, maintains the vital internal parameters such as the body temperature or glucose concentration in the blood. The second is well known reflex. It loops through the environment but works on the same principle. If you notice that your furniture is broken, will try and fix it.

The brain is a control system but also a living organ. As such, it requires maintenance as well. How can it control itself? The idea is the same again. There are 2 blocks: input and output. The insula and other paralimbic areas receive data about brain operation and evaluate its adequacy in the current circumstances. On the other hand, parts of the prefrontal neocortex not only generate abstract plans for motor areas but also can control other brain parts. For example – attention management.

This completely demystifies consciousness as some superability. First, it is only a principle, should be implemented yet. Second, programs without self-control are still quite workable. Finally, adding this feature does not makes the system superintelligent automatically. Elements of consciousness were already implemented in many devices. The frontal cortex keeps programs that generate visible behavior. Adding such a block to a machine means nothing yet. The programs implementing human-like consciousness should be written yet.

Basic cycle

The function of consciousness is supported by several brain structures, but even on the purely functional level, it is distributed. Different elements of conscious self-maintenance are possible. When an algorithm is performed in the real world rather than constant office conditions, it is impossible to foresee all the options, but we can add extra checks. Before the execution, we can check necessary conditions and possible consequences by means of forecasting. When the job is done – evaluate results, draw conclusions, and change some rules for the future.

Machine consciousness

Some elements of self-control were already implemented in computers long ago. For example, a powerful processor generates much heat during intensive computing. Such machines would break down exactly when their workability is especially required. To prevent this, modern models add a temperature sensor which forces the cooler to spin faster when necessary. Consciousness turns out to be not so mysteriously superior as it seems.

How to create a control system which controls itself (among other things)? First of all, what is control? There are 2 concepts. The simplest is regulation, that is maintaining some static condition. A more general one is management, but it may be reduced to maintaining some trajectory, that condition depending on time. Hence, the first variant is theoretically enough. How can a control system maintain own workability?

There are numerous methods depending on a particular device. First of all, we can fix the common shortcoming of the algorithm. Unpredicted circumstances. The environment has changed, but the machine doesn't know and continues the old operation. This leads to errors. We can implement additional monitoring for such situations and methods to fix it. This will require real-time automatic problem solving.

Some programmatic models are already half-conscious. Asynchronous computing is performed not on the time basis but until the result. In many cases, this approach will solve the previous problems. Nevertheless, it has own shortcoming. Suppose the situation is difficult, and the task can't be completed. The whole system will hang, wait to infinity. In this case, a conscious behavior would be to detect the time-out and switch to problem solving as previously.

Computational advantages of consciousness

Consciousness naturally emerges in advanced asynchronous systems. Let's illustrate this on a simple example. In algorithmic computers, the next instruction is retrieved from memory at the next pulse of the clock generator. In a human neurocomputer, the next action is launched by the end of the previous one, but how to detect this end? Keep in mind that we need a universal scheme for all the types of activity. Obviously, the procedure is context-dependent. In the simplest case of one numerical parameter, we can use the primitive threshold function, but still, this parameter itself should be chosen before we can measure its value. In the most general approach, we need to load a special, task-dependent program for this purpose, that is just for running the computational process as such. Already this particular example highlights that we implement self-control, monitoring own actions. That's consciousness.

The same perspective is visible for advanced synchronous computing. We can imagine a device with the changeable clock frequency and, accordingly, a computing system which monitors the current situation and speeds up or slows down when necessary.