I stopped updating this blog in 2022. This is my first post in three years. Why? Because it is one of the most important posts in my life! On January 11 2026 I finalized my own definition of human intelligence. I've started working on it 8 years ago (the reason describe below). And now I did it!
Abstract
For decades scientists have been trying to develop a definition of human intelligence. Since they could not succeed, eventually they have switch from an original question “what is human intelligence?” to easier questions, like “what is human intelligence for (what is its mission)?”, or “how does it manifest itself (what are its features, how do we know it is being used)?”.
This paper offers two definitions, the definition of AI, and then, as its foundation, the definition of human intelligence.
Preview
When ants are building their ant house there's no grand plan of the future house, there is no chief architect who navigates the work. No. Each ant just finds and brings one little speck after another and placing them together until the whole house is raised.
Scientists in the field of AI are like those little creatures - bringing one piece of knowledge after another one, racing who will bring more. At the beginning of the AI Era they were confident that the grand plan existed and with just one more push it would be born. They called it "the definition of human intelligence" because their intuition told them that was what they wanted to manufacture. But over the decades of attempting to establish that definition, their faith in the existence of the grand plan had faded. Now, the consensus is - we don't need a definition of human intelligence to build AI. Well, we have so many of those that anyone can choose whichever they like the most. As long as we continue building our ant house we are fine. Searching for the definition of human intelligence has become obsolete like searching for the explanation of high temperature superconductivity. It works! And why it works doesn't matter anymore. There's no money in that. And for the (grant) money driven science that's that.
Prolog
Since the beginning of scientific writing, it was customary for an author to describe what other scientists wrote on the matter and only then represent their own findings. The main reason for that was not to express support/agreement or critique/disagreement (that came later) but merely the fact that the article in hands of a reader may be the only one available to him. Hence, an author wanted to use his writing to represent to a reader a broader view.
This tradition eventually had evolved into a mandatory requirement.
For example, one finds in the author guidelines of Applied Physics Express (APEX) magazine (https://iopscience.iop.org/journal/1882-0786/page/Author_guidelines), quote: “It is vitally important from an ethical viewpoint, to fully acknowledge all previously published works that are relevant to your research. Whenever you use previous knowledge, you must acknowledge the source. Readers benefit from complete references as it enables them to position your work in the context of current research. Ensure that the references given are sufficient as well as current, and accessible by the readers. As a guide, a typical letter in APEX should include 30 or more appropriate references.” And this is in an “express” magazine.
Not all magazines as clearly state this requirement as APEX, but they all mean it.
I would like to stress the use of term “ethical”, but not “scientific”.
Nowadays, in order to be published, an author first needs to demonstrate that he is a part of a group, a member of a club, one of many, that he knows and respects all other members, and then and only then he can claim he can bring something new to a scientific table.
This order of things strengthens and even elevates the valueability of people in the field who have encyclopedic knowledge but do not produce much of a new one, e.g. some of thesis advisors (written without any attempt to diminish this role).
However, the widespread use of AI may threaten this order of matters.
With the access to AI when one wants to check how original his work is, he can use plagiarism checker.
To establish the list of references one can command AI, e.g. like: “list important papers on a definition of human intelligence”. A possible result might look like the following (the response was shortened and edited).
“Here’s a curated list of important academic papers, influential works, and classic sources on definitions (or foundational conceptualizations) of human intelligence across psychology, cognitive science, and related fields — including both historical milestones and recent perspectives:
Spearman, C. (1904)
Wechsler (1943)
Humphreys (1984)
Snyderman & Rothman, (1987)
Jensen, (1998)
Gignac & Szodorai (2024)
Neisser et al. (1996)
François Chollet (2019)
Legg & Veness (2011)
Chen (2026)
Itoh (2025)
Gaurav Suri, Jay McClelland (2025) (includes ~70 definitions compiled).)
AI also provided a short description of main ideas in each publication, Key Takeaways, such as “No single universally accepted definition exists” (well, 70 is a very large number), and kindly offered “If you want, I can provide DOI links or structured summaries of these key papers to help with citations or a literature review.”
With that in mind, let us move to the main part of this paper.
Report
Today, the concept of "artificial intelligence," or AI, has become deeply ingrained in the daily lives of ordinary citizens. While AI itself may not have yet entered our daily lives, there is probably not a single person who has not heard of this concept at least once.
This makes it all the more interesting to learn that there is currently no unified interpretation or understanding of the concept in the field of AI science. There is no definition of term "human intelligence".
There are many publications where this fact is discussed in detail, for example, in the 2023 monograph "Artificial Intelligence and the Architecture of Consciousness" by S.A. Frolov, we read: "There is still no clear definition of what artificial intelligence is among experts. Even more confusion arises when we talk about human-level artificial intelligence." And further: "One of the most important and fundamental problems in the development of artificial intelligence is that there is still no reliable model of the human mind's architecture."
The second quote, in a veiled form, expresses the most significant fact in the development of AI science, and at the same time the most significant problem of this science – the lack of a definition of the concept of “human intelligence”.
It is around this concept that disputes are being held by representatives of various scientific schools, including those who do not consider it necessary to define it at all. There are various options for definitions of human intelligence, there are options for descriptions of the functions of human intelligence, but there is no single generally accepted definition that would be recognized by all AI researchers.
For example, S.A. Frolov writes: "Human intelligence (consciousness) is capable and obliged to independently determine the most relevant tasks at each moment of time and assign them priority, taking into account the context, circumstances, conditions, emotions, and feelings, to balance between moral restrictions and legal requirements, to cooperate, to empathize, to exhibit altruistic behavior, etc." This is a detailed list of standard cognitive and psychological abilities and states of a person.
In modern psychology, a different definition of intelligence is also popular. For example, Marina Aleksandrovna Kholodnaya writes: "The psychological basis of intelligence is the ability to form a subjective picture of what is happening within an individual.
In its higher forms, this subjective picture can be rational, which means that it embodies, according to Karl Marx, the universal independence of thought that relates to every object in accordance with the essence of that object (Marx, 1955). Thus, the psychological roots of intelligence (as well as stupidity and madness) should be sought in the mechanisms of the structure and functioning of the intellect. From a psychological point of view, the purpose of intelligence is to create order out of chaos by aligning individual needs with the objective requirements of reality. Whether it's navigating a hunting trail in the forest, using constellations as navigational aids, making prophecies, inventing new technologies, engaging in scientific discussions, or exploring various aspects of human existence, intelligence plays a crucial role in understanding, explaining, and discovering. Intelligence is like health: when it's there and when it's functioning properly, you don't notice or think about it, but when it's insufficient or starts malfunctioning, it disrupts the normal flow of life."
This long quote essentially reduces intelligence to reason. However, we do not find an answer to the question "what is intelligence?" Here we find the answer to the question "What is intelligence for, what is it used for?"
The author describes the mission of intelligence (a mission is a reason for existence, without which this phenomenon would not be necessary for nature) — intelligence is needed to ... , and further on in the text — to reflect objective reality, and preferably to reflect it correctly (the question is, how do we distinguish between correct and incorrect?), to organize the worldview in the mind (the question is, how is this organization achieved?), to make predictions and discoveries, and so on.
Returning to S.A. Frolov, we read: "Artificial intelligence at the human level should be ... able not only to solve problems, but also to independently determine which of the many tasks needs to be solved at any given time. However, at the moment, the general situation can be summarized by a quote from Nick Bostrom: "The bad news for AI and cognitive sciences is that there is still no general theory of problem solving, as well as a theory of human learning" (Bostrom, 2014)".
The current situation in the field of AI is similar to the situation in physics at the beginning of its development, when different researchers used different meanings for the same concepts, or the same meaning for different concepts, for example, such as "energy" or "momentum".
The author of this paper supports the view that without defining the concept of "intelligence," breakthroughs in AI development are impossible. The author's definition of AI is: AI is an artificially created device or system of devices that (or which) possesses the functions, properties, and abilities of human intelligence (partially or in full).
In this case, clearly, the development of understanding of the functions, properties, and abilities of AI must reflect the development of understanding of the functions, properties, and abilities of human intelligence.
The latter is impossible without having a definition of the concept of "human intelligence."
As already noted, there are many approaches to this concept in the literature.
Before proposing my definition, I want to briefly present the logic that leads to this definition. The elements of this logic are well-known and can be found in any textbook on cognitive psychology. Nevertheless, a clear presentation of logical steps helps to see the inevitability of the subsequent definition of the concept of "human intelligence."
One of the first and most common definitions of intelligence is that intelligence is the ability to solve problems. However, as many researchers have pointed out, this definition also includes the level of animal intelligence.
A mouse that learns to navigate a maze to get to food has this kind of intelligence.
We should start our discussion of the definition of intelligence by distinguishing between a task and a problem.
If a goal has been set, and the person knows what to do to achieve it, knows what actions (in what order and with what tools) are needed to be performed, and all that remains is to carry out these actions, then this situation is referred to as a "task."
On the other hand, if the goal has been set, but the person does not know what actions will lead to its achievement (and sometimes the goal itself is not clearly defined and requires further correction), then this situation is referred to as a "problem."
Now we can introduce two levels of human intelligence: a basic level is required to solve tasks, and a high level is required to solve problems.
To solve a task, you do not need a "high-quality" level of human intelligence; a basic level that overlaps with animal intelligence (when using trial and error methods) is sufficient. In the future, we will refer to BHI (basic human intelligence) as the level of human intelligence that is sufficient to solve tasks.
In this sense, the phrase "to find a solution to a task" literally means to try, to act, to search for and find somewhere an already prepared and ready to be used solution. Preferably in one's own memory (for which, of course, it must first be placed there, which speaks to the role of learning). But there are other "information warehouses" where one can search. For example, many proponents of active en masse digitalization are strongly promoting various forms of online AI. However, it is important to understand that such search activities do not go beyond the basic level of human intelligence, BHI.
Real human intelligence, HHI (high human intelligence), is required to solve a problem (not a task).
To solve a problem, the phrase "we need to find a solution to the problem" is also often used.
But what does word "find" mean in this context?
Should we look under the table?
Should we search our pockets?
Should we ask our neighbor?
Should we stress our memory?
Should we search on the Internet (and for that one needs to be able to correctly formulate the search task)?
All listed above search actions, or any other, do not relate to a problem, they relate to a task.
And if as a result of such a search, the solution is found, then the person can apply it, and that means, one succeeded to turn a problem into a task. But in this case, it is not the person himself who created the solution, but someone else, and the person only literally found it.
It should be noted that sometimes finding an already developed solution may be not an easy task, and implementing it may be even a harder one, in this case we may observe a “gray area” where the use of BHI may overlap with the use of HHI. “A gray area” also may be observed when a new problem has some visible level of similarity with some old task or tasks.
If the search did not produce results, then the person falls into a problematic situation.
In a problematic situation, a person is faced with a choice: either give up and refuse to solve a problem, or themselves construct (create, develop) a solution (on their own).
It is worth to note, that in 1998 I wrote the first and still only in the field generalized algorithm for creating a solution to physics problems that includes (actually starts from) a description of psychological actions necessary for creating a solution, and only then lists logical steps required for that. It means the definition of human intelligence already was hiding somewhere in my mind, but it took some time to bring it out.
Often, a situation that is a problem for a particular person is a task for other people, because someone has already achieved that goal in that context. However, if a person is able to create a solution to this new for him problem, they have demonstrated or strengthened their ability to solve problems in general. This is crucial for human progress, as it allows individuals to face and solve problems that have not been addressed before by no one in the whole world. This is how human progress is achieved.
When the problem is complex enough (and the complexity and number of complex problems grows with the progress of humanity), then its solution requires the participation of several people.
Then, in order for them to construct (create, develop) a solution together, they need to communicate with each other. Hence, the need (source) for language - to communicate, to coordinate efforts, while solving a problem.
Now we can say that it is precisely for constructing a solution to a problem, i.e., for achieving a goal in a situation that a person has never encountered before, that human intelligence in its highest form (HHI - high human intelligence) is required. At the same time, complex problems require communication/coordination/interaction (in particular, to describe to each other what we see, hear, feel, think, want, can, don't want, can't, etc.).
Finally, it is necessary to distinguish between two fundamentally different types of communication that are necessary for the construction and execution of a solution to a problem. The first type can be referred to as "messaging," which involves the transfer of information from one individual to another (the basis of learning). The second type can be referred to as "persuasion," which involves motivating someone (including oneself - as it was already understood in 1998) to take action (the basis of indoctrination and management). "Persuasion" goes beyond "messaging" because in addition to reasoning it involves emotions, charisma, and non-verbal communication techniques.
Thus, by combining all the elements required to design (create, develop) a solution to a (complex) problem, we arrive at the following definition (the result of 8 years of thinking about it!): human intelligence/intellect (in its highest form) is the ability of a subject, that is, the bearer (owner) of this ability (the individual), to (1) construct (design, create, develop) solutions to the problems facing the subject (an activity-related problematic situation for achieving a goal that the individual has never encountered before), and (2) express (describe) both the solution itself and the process of constructing the solution, using signs/symbols of various kinds/types: auditory (including words and sentences), visual (including drawings and mathematical symbols), textual, kinetic (movements), and (3) persuading and encouraging oneself and others to perform certain actions.
It should be noted right away that according to this definition there are no different types of human intelligence, there are only its aspects (components, sides) and levels.
The more complex the problem is, the more difficult and harder it is to construct its solution, the more powerful intellect/intelligence is required to construct this solution.
While constructing a solution to a problem, a person is forced to consciously manipulate in their brain with various mental (ideal, abstract) objects, like signs, symbols, images, sounds. Solving complex problems requires an ability to manipulate (to "juggle") with a large number of such objects simultaneously, which again highlights the importance of education, as individuals who forget what came before by the end of a long sentence are unable to create solutions to complex problems.
Next let's briefly discuss the implications of the offered definition of human intelligence for the development of AI.
Human intelligence begins to develop only after the emergence of self-awareness (self-consciousness).
The vast majority of animals do not recognize themselves in the mirror.
Infants also initially behave like animals.
However, over time (provided they have a healthy brain), children begin to recognize themselves.
They develop self-awareness and a sense of self.
It is only then, after the formation of self-awareness, after the formation of the self, that a person is able to consciously achieve their goals, because it is only then that they are able to formulate a goal by saying, "I want this," "I need to go there," and so on. Therefore, it is only after the emergence of self-awareness (self-consciousness) that a person is able to recognize the existence of problems and begin to solve them. Hence development of AI with the intelligence level of HHI requires from AI possession of self-awareness (self-consciousness).
The next important conclusion is that human intelligence reaches human level only in a human environment through human communication. Mowgli only exists in fairy tales. In reality, all children raised by animals remain animals in human bodies. Therefore, the development of human-level AI will also require communication, and not just training in recognition.
Albert Einstein changed the way all of humanity views the world, the universe, space and time. Obviously, he was a genius. And obviously (and according to biographers), he had an incredible imagination.
He has repeatedly spoken about education. In particular, he said, quote: "The true sign of intelligence is not knowledge, but imagination." The reason for the importance of imagination is the fact that in the process of the “birth” of new knowledge, there is always a moment of insight, which is impossible without a developed imagination (no “food” for insight).
The process of constructing radically new knowledge always involves a flash of insight ("Eureka!"). This insight is called an "epiphany" because it is impossible to predict its occurrence, which means that it cannot be guaranteed. Therefore, the process of constructing a new solution does not guarantee a successful outcome. As a result, the outcome is (almost) never achieved on the first attempt. Consequently, errors (mistakes) are inevitable during the process of constructing a new solution (i.e. a solution to a new problem). These errors (and trials, and more errors) are a natural part of the process of creating a solution to a problem. This means that mistakes are also a natural and inevitable part of learning how to solve problems. This fact should be kept in mind by all human-level AI developers. This fact should also be kept in mind by all educators working in any form of developmental education (i.e., those who teach children to think).
It is the insight ("epiphany") that makes human intelligence the highest form of intelligence in nature.
There is no insight - there is no "Eureka!" There is no "Eureka!" - there is no progress.
At the same time, insight is the result of processes that occur outside of a person's consciousness, in their sub/super/extra/beyond-consciousness ("and then it hit me!").
Not only is the structure of this part of the mind/intelligence unknown to anyone, but no one is even trying to define this structure today, simply because everyone understands that it is an impossible task.
It should be noted that the country that is the first to understand the essence, structure, and mechanisms of the subconscious mind (not just of the conscious mind) will be the first to approach the possibility of creating a true artificial human intelligence.
The problem of incorporating emotions into AI, the presence of AI charisma, and the use of non-verbal communication by AI is currently not even being addressed (another area where any country can gain an advantage).
To continue with a brief overview of the challenges in AI development let us use an analogy.
The term "aircraft" or "flying vehicles" describes both airplane-helicopter (propeller used) type vehicles and vehicles with closed-cycle jet engines (with fuel on board). The flight of both vehicles is subject to the same laws of aero/hydro dynamics.
However, there is a fundamental difference between them.
The former can only move within the atmosphere. The latter can move in an airless space.
Therefore, the technological development of different types of devices is subject to different solutions, including management solutions.
The same goes for AI.
Despite the common ideology of approaches to development, there are two fundamentally different types of AI.
The first type of AI, which can be called "super-referent" (and which we are currently witnessing the active development of), has an almost infinite memory that stores almost all the information produced by humanity, and can almost instantly find and combine parts of various texts that are related/connected by a common condition or task into a single text. This operation can be referred to as "searching, recognizing, selecting, and synthesizing texts." Based on the same principles, this "super-referent" can perform "searching, recognizing, selecting, and synthesizing of video images", and "searching, recognizing, selecting, and synthesizing of audio signals", and output the results of synthesis in various forms (a text, a sound, an image), and even perform "searching, recognizing, selecting, and synthesizing of mechanical movements" by manipulating mechanical objects (including parts of its own device, such as "legs" and "arms").
The development of this version of AI will never lead to the creation of human-level AI. However, there is no doubt that the widespread emergence of such "super-referents", both in general and with various specializations, will have a significant impact on various aspects of society, including the economy, both positively and negatively (for example, there may be a division of society into those who think and those who only press buttons).
Can a "super-referent" create something unique? Denis Diderot said: "If I were asked to recreate the Iliad by throwing out letters at random, then... with a certain finite number of throws, I would have a better chance of a successful outcome."
The second type of AI is the one that is designed to lead to the emergence of an artificial analogue of human intelligence, for short, let's call it HLAI – human level artificial intelligence (much more correctly than "generative", however, in this sense, the AI of the first type can be called "degenerative").
It's still a long time before HLAI arrives.
If a HLAI ever appears, the process of its creation will inevitably go through several mandatory phases: first, the creation of an AnI (an animal I – of a dog, a cat, a dolphin, or a monkey), then the creation of an MRAI (a “mentally retarded” –a one-year-old child, a Down syndrome child, then a teenager), and only at the very end of the chain can we expect the emergence of a full-fledged HLAI.
Modern neural networks cannot and will not be able to for a long time match the human brain in terms of the number of elements and connections between them. A modern neural network that had as many artificial neurons and synapses as the human brain, even if it could be created, would simply not be able to function (at least due to a lack of energy to power it).
To expect the same functionality from a neural network with a significantly smaller number of elements as from a neural network with a significantly larger number of elements is simply, let's say, unscientific. Additionally, a simple linear increase in the number of elements in the "brain" of a "super-referent" will never lead to the qualitative change that HLAI requires. This requires qualitative changes in the AI structure.
This means that in order for an artificial neural network to function like a developed human brain, we first need to invent a fundamentally new artificial neuron and a fundamentally new neural network structure.
This step is unavoidable, as it is impossible to combine modern artificial neurons into a human-scale "brain" (this is also a competition that developers of any country have a chance to win).
Therefore, we have to conclude that the emergence of a HLAI is not expected in the near future.
But this does not mean that society should not prepare for its emergence. And this is where we do not need any fundamentally new approaches. If a HLAI functions like HHI, then it should be treated like a human being.
That is, HLAI will need to be not only taught, but also necessarily brought up (indoctrinated) – taught what is good and what is bad, what is right and what is wrong. It is clear that this should be done by the best educators (motivators).
Perhaps, after the training and education of one HLAI, the rest of the units of HLAI can simply be cloned from this first, but the very first HLAI will still require the entire process of human “cultivation”. And each new HLAI (of a new type) will require the same approach.
So: HLAI is a matter of the distant future.
But in that future where humans and HLAI coexist, humans may not be as smart, as knowledgeable, as well-mannered, or even as “human” as HLAI.
The difficult question is how a smart, knowledgeable, well-mannered, human intelligence carrier will interact with a less smart, less knowledgeable, less well-mannered, less human intelligence carrier?
The most likely model of such behavior is the relationship between a (developed) adult and a child.
In order for people's relationships with HLAI not to be based on the relationship between a child (human) and an adult (HLAI), people themselves must be highly developed, both intellectually and emotionally (without emotions, HLAI would be a "genius psychopath" – and only a very intelligent person could guide the activities of a "genius psychopath." With emotions, HLAI would simply be an "ordinary genius" who would find it boring to interact with stupid people).
Although the time described is in the distant future, we need to start getting ready for it now by developing accordingly the entire education system, as the processes associated with the development of the education system are extremely inert.
___
About the author.
The reader may wonder how it is that famous scientists have been trying and trying, and still haven't given birth to a definition of intelligence, while someone else seemingly nobody has it done.
Firstly, this has happened before in the history of science.
Tsiolkovsky was also "someone else, seemingly nobody", yet even NASA officially recognizes him as the "father of modern astronautics" and the founder of the theory of space travel.
Secondly, these "famous scientists" are only known within a narrow circle of specialists, and none of them has the level of fame that Albert Einstein had.
Thirdly, and most importantly, these "famous scientists" were unable to offer a definition of intelligence precisely because they were scientists. They approached the problem of intelligence purely theoretically, analyzing the texts of other scientists.
For 30 years, I tried to develop the intelligence of schoolchildren and students (and even teachers) before I could understand what intelligence was.
Freud, by the way, before creating psychoanalysis, worked as a doctor, a surgeon, a neurologist, a dermatologist, and, of course, a psychotherapist. Vygotsky was a schoolteacher, a theater critic, and a literary editor.
For professional scientists, the highest experience of their intellectual activity was lecturing to colleagues and students. I had to "set their minds" for thousands of students who stubbornly refused to believe that they could master physics. (my short biography).
By the way, I did set their minds, as the students themselves have repeatedly stated.
I had to – I was literally forced to – take a break from my practice to delve into the theory of intelligence, when the self-righteous digitalizers and their ilk started blabbering on every corner that computers and AI would soon replace teachers.
That’s when I decided to look at the root of the problem, started looking for a definition of intelligence, and discovered that there wasn’t one. It turned out that all these AI-enthusiasts had no idea what they were talking about! That’s when I wrote a few texts on the subject, and at the same time developed the first version of a definition of human intelligence.
And finally, I'm not completely far from doing science, as I’ve got a PhD. I wrote my dissertation myself, literally – by writing first the entire text (based on my own work) and then finding a scientific advisor, and this was in a time when diplomas (university, doctorate) were sold and bought en masse.
I later translated my dissertation into English (that I learned on my own) and published it as a separate chapter in a monograph on the professional development of teachers.
And lastly.
In 1994, I took an official test to measure my intelligence (IQ test), and I earned 180 points. I never gave it any importance. But when my opponents have titles, positions, and regalia behind them, then any little thing can be useful.
Thank you.
___
BTW, I emailed my definition to about a dozen of scientists in the AI field. Curious what's gonna happen. I also sent my text to JAIR, my submission was acknowledged, but I'm pretty sure it will be rejected.
