When perceiving external information, one must be aware of three magic letters that represent the processes into which this overall process of perception is broken down. These letters are P, R, and S, which stand for the words “Parsing,” “Recognition,” and “Structuring,” respectively.

In psychology, the description of the process of perceiving external information is conventionally divided into perception and apperception – the primary sensory perception and the logical level that follows sensory perception.

However, this division is rather arbitrary and is not suitable for the algorithmic construction of perception. Therefore, three magic letters were introduced to make the process more understandable. Moreover, these letter-processes are so closely related that they are intertwined and can appear one after another many times.

For example, when we parse a text, we separate words during perception. Some very long words have to be structured (parts of the word are extracted) and templates (which relates to recognition) are applied based on existing words. And this process can continue.

The result of recognition should be a congruent (related to your existing concepts) and structured result, for example, in the form of a tree, of images, words, or other elements – primitives or granules – used in the description.

Let’s discuss code autogeneration, starting with a description of Russell’s paradox, well known in set theory:

Let K be the set of all sets that do not contain themselves as an element. Does K contain itself as an element?

This paradox is typically illustrated by several equally amusing and absurd problems:

The Barber’s Paradox: A village barber is ordered to shave anyone who does not shave himself, and not to shave anyone who does shave himself. Can the barber shave himself?

The Mayor’s Paradox: A country decrees that mayors must live not in their own city, but in a special city for mayors. Can the mayor of a city of mayors live in a city of mayors?

Russell’s Paradox is undecidable due to the very formulation of the problem, since it refers to the set of all sets that do not contain themselves. In other words, the contradiction in the problems is axiomatically defined, in the very definition. To bring the mathematical view of things to a natural, and, if you will, to a common understanding, let’s compare this paradox to the omnipotence paradox: ask an omnipotent being to make a stone they can’t lift. If they succeed, their omnipotence has expired; if not, they were never omnipotent.

A similar paradox, related to modifying program code, arises in artificial intelligence problems. Specifically, can program code modify and construct itself? Or, in another interpretation, can program code construct other program code that doesn’t contain itself?

How do the sets in Russell’s paradox relate to the generated program code? At the most basic level, code consists of commands and operations. Their orderly sequence is the code that needs to be produced. But at the level of the generation problem statement, there is no program code; it is generalized information, or meta-information, about what the generated code should be. This is a set (of tasks) that does not contain another set (of commands). A task is an instruction to the processor to execute a specific instruction. It turns out that by creating a program and writing it to memory, we instruct the processor to execute a specific set of instructions. As the processor follows this sequence of instructions, it considers each one in the current execution context (registers, memory state, etc.). The concept of a set that does not contain itself is essentially meta-information (tasks) about meta-information (commands), which does not contain lower-level tasks, that is, more detailed representations of tasks, although formally this more abstract form of meta-information remains meta-information.

I’m not sure everyone can pass the Turing Test. This uncertainty can plague anyone who’s spent enough time in chat rooms. If the person on the other end of the line has difficulty verbalizing (expressing thoughts), knowing the language, or perceiving (perceiving), then we, if we conduct a Turing test – that is, if we doubt that the person on the other end of the line is human – might give them a negative result. They won’t pass the test. It can be concluded that the Turing Test isn’t a test of intelligence, but a system for recognizing “friend or foe.” In other words, it’s a test of whether the person they’re talking to matches the tester’s concepts and standards.

The only conclusion to draw from this difficult situation is to reject the concept of “intelligence” as a goal toward which artificial intelligence should strive. There’s no need to measure intelligence as a single, indivisible whole, just as there’s no need to try to define the criteria of intelligence, since it’s an empirical concept comprised of hundreds of characteristics.

What, then, must we actually do if we still wish to use a machine as an intelligent mechanism? Obviously, we must identify the components of intelligence that are valuable both for humans and for their alienation to machine implementation.

There’s a seditious idea floating around in the field of artificial intelligence. It goes something like this: “If we replicate the human brain, we’ll get intelligence.” Norbert Wiener encourages this notion: “…cybernetics assumes that the structure of a machine or organism is an indicator of its ability to perform a task. … the mechanical rigidity of an insect limits its intelligence, while the mechanical flexibility of a human being ensures its almost limitless intellectual development… Theoretically, if we could create a machine whose mechanical structure replicated human physiology, we could have a machine whose ‘intellectual abilities’ would replicate those of humans.”

As if intelligence were the result of a chemical reaction or a soup brewed from precisely calibrated ingredients according to some mythical recipe. Unfortunately, little can be achieved by blindly copying and waiting for subsequent synergy, or the transformation of quantity into quality. It’s impossible to expect the created structure to fill the gap in understanding of a whole layer of knowledge and process, and to begin reproducing “intelligence” on its own. Miracles won’t happen. One reason for this is that, in addition to reproducing mechanisms, these mechanisms must at least be imbued with knowledge (innate or a priori reflexes, instincts, etc.). I believe that if such an organism could be created artificially, it would be similar to a clone of the human being. But do we really need a clone? Is this what we were striving for when creating such a mechanism?

One of the problems with the concept of “machine intelligence” is that, under the influence of intelligence tests (like the Turing test), “artificial intelligence” becomes synonymous with “human emulation.” Moreover, machine intelligence is expected to accurately replicate all human traits, including errors, inaccuracies, and inabilities, such as the human inability to calculate quickly and accurately.

But do we expect artificial intelligence to fully emulate human capabilities? Obviously not. However, this goal of creating a complete analogy between machine and human is a leitmotif in many artificial intelligence technologies. For example, the idea of ​​replicating the brain gave rise to the development of neural networks. The desire to “grow” an ideal, self-developing structure gave rise to evolutionary algorithms and the game of life.

At the same time, copying the internal mechanisms inherent to humans and replicating human capabilities are different things. Perceptrons, reaction processing, and other mechanisms help us best realize certain ideas and tasks, such as pattern recognition. Conversely, completely replicating a human being is futile, since blindly copying the result lacks key value: a human being is merely a vehicle, one embodiment of the intelligence we strive for in software systems.

A computer has its own important properties that make it valuable. These include speed and accuracy of calculations, rapid data retrieval, reliability and precision (discreteness) of logic in executing program instructions, and the ability to store large amounts of well-structured data.

Human thinking differs significantly from the rules of computer data processing. It lies in the ability to make decisions under conditions of limited information, the lack of clearly structured information, and the dynamic construction of operations. And, generally speaking, can a computer and a human being be compared? Complex concepts such as sensuality, intuition, dreams, imagination, expression, and spirituality cannot be applied to a computer. But this doesn’t make a computer any less capable of calculations and program execution. This doesn’t make us work with it any less. So let’s consider artificial intelligence, in its computer implementation, not as a competitor or a brother, but as a tool.

The theory of multi-agent systems (MAS) is best suited to examining the “society” of intelligent systems. This is a system formed by several interacting intelligent agents. An intelligent agent is a subject that observes or acts within its environment. Such an agent can be a robot, a software system, a human, etc. Communication is crucial, but since we are studying computer systems, we will pay more attention to agents as software systems.

Agents are divided into agents with simple behavior, agents with model behavior, goal-oriented agents, practical agents, learning agents, etc.

In MAS theory, subagents are distinguished separately. A subagent is a part of an agent that can be isolated into a specialized subsystem. Thus, there are:

• temporal subagents for making operational decisions,
• spatial agents for interacting with the real world,
• learning agents, etc. Subagents can serve various purposes, and they are largely categorized based on the processes and architecture of the agent itself.

Agents in a multi-agent system must have several important characteristics:

• autonomy;
• limited visibility. That is, no single agent has a view of the entire system;
• decentralization; that is, no agent controls the entire system.

In a multi-agent environment, individual agents have the ability to receive and generate information fields in the form of messages between themselves. All agents are not necessarily equal in roles, rights, capabilities, interdependencies, accessibility, or openness to interaction with other agents. For example, in a multi-agent environment, alongside intelligent agents, there may be data-sharing agents, such as search servers and public repositories for data exchange and retrieval.

A system must be able to continuously evolve and must be capable of universal task performance. Universality refers to its ability to integrate into the overall ecological system of its existence (society).

Universality is the most valuable property of intelligent systems. It signifies the system’s greatest flexibility, its ability, on the one hand, to capture, perceive, recognize, and assimilate external data, and, on the other, to generate responses, reactions, information, and actions that fit within a generally accepted knowledge structure. A generally accepted knowledge structure is a system of concepts, terms, methods, and theories accepted in society.

The system’s fundamental properties are its communicative capabilities within society. In other words, it is not some computational function or an intellectual mechanism that is being considered, but the ability to communicate within a shared environment. At the same time, a single element of a system possessing only communicative properties is degenerate, as it carries no value associated with the retrieval, storage, structuring, and analysis of information.

However, a “mind within itself” that lacks sufficient communicative capabilities is also a degenerate system. At the same time, it is impossible to maintain an acceptable level of communication without information processing. Specifically, a system will be unable to answer questions without an information storage and retrieval system.

It’s an unpleasant question, isn’t it? How can anyone even think that some machine is better than nature’s crowning achievement? But let’s put that aside, and be less arrogant. Let’s look at this question pragmatically. We’re not measuring strength, but rather learning how machines can help us.

So, compared to humans, a computer has properties that are beyond humans’ control:

– data memorization accuracy and long-term memory. Over time, humans are capable of forgetting information, including important information. A computer stores data exactly as it was saved.

– data storage capacity. A computer can store a wealth of data that is impossible for a human to remember. For example, in Human Resource Management Systems, computer systems can store data on tens of thousands of employees who have ever worked for the company. No human being could do something like that.

– ownership. A computer is owned, meaning a specific person and company can be confident that the data will remain theirs.

– accessibility. Information stored on a computer is accessible 24 hours a day, 365 days a week.

– reliability. As one of Murphy’s laws states, “Computers are unreliable, but people are even more unreliable.” This property follows from the principles of availability and certainty of data structures.

– certainty of data structures. A computer system should allow the user to analyze itself. The computer’s internal mechanisms should be understandable and transparent, that is, analyzable. In contrast, the form in which human knowledge is stored is uncertain. The mechanisms by which human knowledge is stored are opaque and inaccessible to outsiders.

The prevailing belief is that humans use the decimal number system because they have ten fingers.

As you know, computers use the binary number system, and at various times, they have used and continue to use operations based on powers of 2: 4-base, 8-base, 16-base, and sometimes even 32-base and 64-base codes. And converting to and from the decimal system is a computer’s least favorite exercise.

Let’s get back to humans: okay, we have 10 fingers, but that doesn’t really tell us anything. Because if we use fingers in a non-positional numeral system – unary or unitary – and each extended finger (no matter which finger or on which hand) represents 1, then the sum of 10 fingers equals 10: from zero (all fingers curled) to 10 (all fingers extended).

In that case, the number of digits in our number system should be 11: from 0 to 10, right? And then we should use the 11-base numeral system.

But it so happens that for some reason we don’t use the 6-base system (the number of fingers on one hand plus zero), nor the 11-base system (10 fingers on both hands plus 0).

It’s impossible to understand this because it’s impossible to know (and that’s why it’s called transcendental knowledge), so we use what’s passed down to us as tradition.

Over the past 70 years (its formal history began in 1958 with the invention of the integrated circuit by Jack St. Clair Cilby), technology has given us microelectronics, which expands its data storage and processing speeds every year.

And now we are witnessing a discontinuity, with computer hardware offering vast capabilities and power, continually growing and expanding according to Moore’s Law (processor performance should double every 18 months due to a combination of increasing transistor counts and increasing processor clock speeds). But computer programs are essentially analogous to complex devices, nothing more.

Every step and every operation is strictly deterministic (defined) and pre-programmed. They were developed, compiled (cast in bronze), and now all we can do is admire them and execute them exactly as they were designed. We can only get an update in the next release, if the development team does a good job.

In this context, let’s ask ourselves: is it possible to isolate human thinking and implement it in a computer? It seems so. At least, I’d like to. But how?

A product’s value lies not only in its consumer properties but also in its environment – the product’s selection and purchase, its brand and creator, the status the user acquires from owning or using the product, its packaging, expectations upon receipt, after-sales service, and much more.

Consumer properties are expressed in the user experience (UX). Everything else described above, and even more, is reflected in the customer journey map (CJM).

And often, the latter is more important than the former: you know, packaging sometimes costs more than the product itself, while a well-known brand with a high price tag has a pittance in production costs.

Financial management defines the goals of any company quite simply: 1) profit, 2) sufficient cash flow to pay obligations.

But what is profit? It’s a financial result. It’s usually preceded by significant work, which results in some benefit to other people or organizations.

What generates financial results? The sale of Deliverables – products: tangible goods, work, services. These are managed by a Product Manager: he or she formulate their content, value for customers, price, market position, and much more.

Financial management defines the goals of any company quite simply: 1) profit, 2) sufficient cash flow to pay obligations.

But what is profit? It’s a financial result. It’s usually preceded by significant work, which results in some benefit to other people or organizations.

But what generates financial results? The sale of deliverables – products: tangible goods, work, services. These are managed by a product manager: he or she formulate their content, value for customers, price, market position, and much more.

The role of a product manager is often blurred: the role can be played by the CEO, chief technology officer, chief engineer, or even distributed among various positions. But it definitely exists: someone in the organization clearly owns the product idea and the value it delivers to the market.

One of the main questions of philosophy are the questions of foundation: “What is all this for?” “Why is it this way?”. Let’s not delve into the depths of the issue (there’ll be time for that!) – ideas of objects, reality, truth and much more. At the very end, we’ll see one very simple thing: value.

Yes, people are very mercenary beings. Despite some of us talk of altruism and nobility, value ultimately underlies interests, principles, and goals.

And in computer science, too: symbols, machines, programming languages, technologies, automation, software and hardware solutions – at the forefront, they produce results that must possess some clear value. Whether it’s the acceleration of processes and calculations, aesthetic goals, the desire for possession and ownership, or the opportunity to embrace modern trends and fashion.

I created a Telegram channel dedicated to the philosophy of computer science.
Join the channel:

https://t.me/infosophy

Since my site is a content hub, the posts will be available here as well.

Philosophy of Computer Science? Something vague, abstract, science-intensive, about everything and nothing at the same time?

Absolutely not. It couldn’t be more concrete. It emerges when we need to look at the same facts and technologies from different angles.

When we need to connect facts within a discipline and across disciplines.

When we need to generate new ideas and rethink existing approaches.

Let’s look at specific cases, provide a forum for reflection and new ideas.

Happy holidays to you – Merry Christmas and Happy New Year.

Just yesterday (at the end of December 2025), I wrote a short prompt in the new version of DeepSeek, and it turns out the new version of the neural network now produces amazing texts – in terms of content and style, complexity, and vocabulary.

Just then, I was writing the next chapter of my book. I had just one question: why would you find it useful to visit my website? After all, AI can answer any question and teach you anything.

The answer: in the modern era of AI, several things remain important, and this website is dedicated to them:

1) Provide a consistent and holistic view of things, not a fragmented narrative.
2) Provide a personal perspective – a point of view, a way of looking at things.
3) Describe questions on a topic that need to be asked – to yourself, to others, to AI. After all, questions are the foundation for thinking and for any other endeavor.
4) New Ideas

Brevity is the soul of wit, so I’ll leave it at that and invite you to explore the content.