How to create artificial intelligence? The second story. Algorithms for intelligent search and storage of information

In a series of articles we talk about new approaches in artificial intelligence, personality modeling and processing of BIG Data, which are not available for most AI specialists and the public. The value of this information is that it is all tested in practice and most of the theoretical developments are implemented in applied projects.



All of you, without exception, use modern search engines that do a good job of searching for links by keywords. But can these technologies be considered a search for information, can they answer your questions in different contexts, can they reflect, imagine on the basis of hundreds, thousands of sources in different languages?
')
Large search companies have almost unlimited resources - they have money, the most professional employees work for them, thousands and even millions of servers serve search queries.
So why are you still getting the link as an answer and not the answer to your question?

In this article we will talk about technologies and algorithms that will change your understanding of the search and storage of information.

In the previous article, we introduced the basic terms, namely - intelligence, learning, imagination, memory, image, information flow, information field. Be sure to check it out !: geektimes.ru/post/249042 .

The article caused a certain resonance, and we received three categories of questions:

• “Is it possible to write an algorithm that could think, think?”
• “Do we really have competence in AI?”
• "What practical applications are developed by us?"

Today we will talk about the algorithm for finding information on the Internet. In order to distinguish it from our other algorithms, let's call it: “core v2.3”. How much the algorithm meets the criteria of intelligence you decide.

The basis of our algorithms is the assumption that the processes occurring during reflection have an electrical nature . It can be assumed that when a person says something (reflects in the ear), different parts of the brain are activated, and they are activated, not simultaneously, but sequentially. Different parts of the brain are responsible for the different stages of the “speech of thought”. It can be assumed that the various "mental aspects" are dependent on each other.

Reflection (thought) is a kind of presentation of the sequence of information units (including the translation of images).

Considering that this assumption is true, we decided that the process of thinking can be modeled on a computer.
Consider an example:
"I will go to you" - this is an image-information structure , it consists of three images:

• “Image of self-awareness” - “I”,
• "Action" - "Go",
• “Comprehensive or unifying image of what we want to achieve” - “to you”.

Thinking in different situations can have different implications (hidden meaning):
For example, "I want to drink tea" in different contexts may have different meanings.

• Who says this phrase - "Woman speaks to man", "Woman to woman",
• At what time of the day “in the evening”, “in the afternoon”, “in the morning”.
• What is the situation “during a walk”, “in the office”, “during a date”, etc.?

“Understanding the meaning” of thinking is “the ability to read between the lines,” from an informational point of view, this is a search for complex correlations (search for informational and structural interdependencies).

We also proceeded from the fact that any text is a kind of logical sequence that was somehow generated by a person during reflection.

We decided to take the finished text on the topic "How to easily quit smoking" and look for correlations in it, but found nothing. We assumed that the training was due to data redundancy, and we have already taken 500 articles on the topic “How to quit smoking easily”. Using various non-linear algorithms, we found complex correlations.

In particular, we have found points at which “lines of application reflections” can be directed along different paths. We call these points “pivot points” or “Motivators” . We mentioned the "Motivators" in the previous article; for this we used a more popular term from psychology - anchor.

A “motivator” is a structural element of a chain that has “great resistance” and allows for a change in the subsequent direction of movement of a signal (thought) under the influence of additional factors (motives) . Thus, a motivator can change the path of “thinking” if “additional potential” appears in the process of thinking.

Another important property of “pivot points” / anchors / motivators is their use as an “entry point” or “moment of thought”. In this case, the train of thought turns out to be consistent. And if you start thinking with another concept (NOT from a motivator), then the first thing that happens is alignment with a motivator. In life, this is manifested as follows. If you are asked to reflect on something unexpected for you, then most likely you will start your reasoning "from far away". When thinking, you have to reach a certain point (to the motivator), and only then continue your reflection in the “given” direction.

In order to process and store such information structures, it was necessary to create a special memory model, we designed it and called it “factor memory”.

“Factor memory” is the structure for the formation of an information field from an unstructured information flow, as well as the definition and management of motivators. In other words, the factor memory model is a memory that uses semantic representations of an image as a set of characteristics - factors, and chains of ordered data - facts.

We remind you that from our point of view intelligence is the ability to learn and imagine.

Let's ponder whether the described core model possesses these qualities?

Self-study, learning objectives, learning criteria in the context of the AI ​​model.

Training is the transformation of information flows into the information field.
The model has a main motive - to solve the problem, to cope with the difficulty coming from outside.
Its task is to recognize the images that come in the form of a request, to create a complete image, using redundant information located on the Internet.

Since the model's memory is clear before the request, it needs to “learn the language” (the language is not clearly defined), to understand the meaning (find complex correlations and motivators), to form a redundant information field in order to be able to reflect (create information chains).

The main technical criterion of “filling the memory” / “creating the information field” / “creating a complete image” is an assessment of the entropy of the information field . Entropy is the criterion of the sufficiency of the image being assembled, and at the same time the main criterion of the lack of information.

Training is not possible without feedback.
The feedback in the v 2.3 model is as follows - the user can adjust the “point of beginning of the reasoning” and choose the direction of forming the answer (reflection) within the framework of the generated information field. The model demonstrates the result of thinking to the user and waits for a correctional clarifying word (motivator).

Imagination in the context of the AI ​​model

Imagination, we defined as the ability to complete the unfinished image on the one hand and the ability to simulate a situation on arbitrary conditions, on the other.
Algorithm, capable of activating different motivators, receive different chains of thought.

Examples of the v 2.3 kernel model

1) Request: "The impact of violent computer games on children."
Context, specification: Without specification.
Thinking model:

“In the spring of 1999, Bill Clinton stated in his televised address to America:“ We must think twice when it comes to “first-person shooters”. ”
They started shooting already in the parking lot, then rushed inside.

Hot on the heels of the April issue of the journal "Personality and Social Psychology" in 2000 appeared the article "Video games and aggressive thoughts, feelings, behavior in the laboratory and in life."

The authors, psychology professors Craig Anderson and Karen Deal, argued: children see more and more cruelty in the modern world.

The article argued that there is a link between the brutality of computer games and the aggressive behavior of adolescents, which can lead to crime and murder.

Anderson said: “Although there are many difficulties in the study of behavioral characteristics, one simple and clear thing everyone needs to know: computer games increase cruelty and violence.
Our study showed that the impact on the aggressiveness of behavior in violent computer games is even more serious than in cruel television programs and films. ”

And what if we take a step back from our conditionally complicated vegan situation and think about who these kids playing into death will grow into.

Many experts disagreed with such an opinion, believing that the conclusions of Anderson and Dil were more of an opinion than a proven scientific fact.

On April 26, 2002, Robert Steinhoiser killed 17 and wounded 7 people at the Guttenberg Grammar School, Erfurt, Germany.

The authoritative newspaper Frankfurter Allgemeine Zeitung published an article in the “Program for the Slaughterhouse,” which said: “The killer trained with the help of computer games.”

The document expressly prohibits "the distribution of any types of products that promote violence, aggression and anti-social behavior."

The Department of Youth Affairs of the regional administration was appointed responsible for the elaboration of measures.
The goal of all these events is to fight youth aggression and violence.
These events provoked a reaction in the neighboring territories of Ukraine, where, as a matter of urgency, a similar Belgorod decree on combating the propaganda of violence was taken, and law enforcement officials began to confiscate "violent computer games" in clubs.

Hard computer games rather discharge mental energy accumulated during the aggressive self-affirmation of adolescents in everyday life.

At the same time heightened activity and aggressiveness of a teenager in relation to the world around her gives him support to overcome his own fear.

Honestly, I myself once for a long time really loved all sorts of shooters, while I did not notice the aggression. "

The request can be clarified with a word or part of a word. For example, “undergrowth”, “games”, “cruel”, etc., after clarification, the thinking will change.

Motivators of the information field:

years --- play --- law ---
ruling --- fight --- propaganda ---
brain --- that --- respond ---
frontal --- share --- person ---
computer --- games --- help ---
computer --- games --- parents ---

2) Request: "Problems of artificial intelligence . "
Context, clarification: "problems . "
Thinking model:

To isolate the structure and understand the principles of the organization of the intellect is to reveal the real basis of the fundamental nature of the problem, to show its deep specificity.

In other words, it is necessary to study the historical, scientific, technical and epistemological aspects of the problem of artificial intelligence with the direct participation of dialectical materialistic philosophy.

Philosophy, which has chosen the problem of the relationship between being and consciousness as the main issue of its scientific research, should be more closely interested in issues related to modeling, reproduction of the intellect.

The problem of artificial intelligence is directly related to the basic philosophical question: it is about artificial reproduction, modeling of human intelligence.

In addition, already the first work on artificial intelligence showed that not only the area of ​​problem solving is affected by comparative studies, but also the problem of thinking in general.

In different types of animals we find nervous systems of different complexity, and if we gradually move along such a ladder of complication in the process of evolution, understanding what mechanisms were responsible for one or another leap in the intellect of animals, then perhaps this will allow us to solve the problem faster understanding of human intelligence.

Motivators of the information field:

Besides --- solutions --- problem ---
Facebook --- Research --- Research ---
human --- activity --- activity ---
Besides --- problem --- solution ---
areas --- artificial --- intelligence .---
problems --- artificial intelligence --- intelligence ---
areas --- artificial --- areas --- artificial --- intelligence .---

3) Query: "how to find your meaning in life"
Context, clarification: "life"

Thinking model:

“In his theory, Frankl dwells on people who are addicted to alcohol and drugs, who do not see not only the meaning of life, but also real life in general.

What is the meaning of a certain person's life?

Has he washed away life for all people on earth or is he unique for everyone?
And if you didn’t manage to enjoy, then you have lived life meaninglessly.

In this regard, it is important to remember in any life segment that our path has not only a beginning, but also an end.

Everything that has been done in your life is important, because a person is remembered long after his death.
The meaning of human life is to continue the development baton from the primitive mind of microorganisms to the super-intelligent creatures of a distant future.

But the most interesting question, the answer to which we will never know: the meaning of life in general.

Happy people do not ask such questions, they just live and enjoy life itself, therefore such questions come to mind when a person is unhappy.

The meaning of life is the same for all living beings, from a microbe to a human: live on its own and leave offspring.

And even cooler - life is just a very complex chemical reaction of mainly carbon atoms flowing in nature, regardless of you.

The purpose of life is in life itself, realizing its divine essence!

Contrary to popular belief, people's lives are ruined not by low self-esteem but overestimated, ambitions flow from it, in the absence of implementation of which a person begins to feel suffering.

The meaning of life in all living things is the same: DEVELOPMENT through the distinction between good and evil, the benefits of harm, the true knowledge of delusions.

As well as the purpose of everyone in this life is different.

You must listen to your feelings, not your thoughts and understand what you want to do in life.
To do this, you need to get away from everyday fuss and stop looking for the meaning of life.

What is incorporated and in what direction in life to be realized.
Having found your vocation, you will begin to enjoy life and every moment.

But life has returned to the original. "

Motivators of the information field:

yourself --- this --- life .---
find --- yourself --- yourself ---
meaning --- life --- life ---
life --- each --- life ---

Algorithm performance

Our research began more than 5 years ago and took place on a specialized cluster. After certain patterns were found, it was necessary to optimize the algorithms. It took several years for the “reflection” to work on the usual server hardware. The kernel version 2.3 can even think on a tablet with an Intel Atom processor.

The 2.3 kernel model is a single memory cell (the “Unicellular Memory” algorithm), one information structure that has sufficient redundancy for an image to be defined within its framework, and present it from different points of view. We have posted the prototype of the 2.3 kernel model on the Internet for testing.

Features and Development of DES Algorithms

If we use memory from several cells, then we can build more complex correlations, in such models there are transitions from one memory cell to another. Therefore, such algorithms are called deductive or deduction or DES. Unfortunately, the public does not know much about these algorithms, since such studies are usually of a closed nature.

Any data stream can act as an information flow: text, video, telemetry, stochastics, etc. The algorithm is able to find correlations and build ordered data both in and without chronology. The use of text data allows you to simply get the information flow and clearly demonstrate the results of the algorithm.

If you want to independently test the capabilities of the kernel v2.3, then request a link from us through the dialogue mode. We do not post the link in public access, because the kernel model is installed on a weak server, and we would not like to give direct advertising.

In order to introduce such technologies en masse, it is not enough to have them, it is also necessary to change the information paradigm, and create new approaches to information processing and storage.

In the next article, we will discuss the technologies of the next generation of the nucleus, this is a multicellular memory model that allows you to model some aspects of the personality, to conduct a dialogue within the user's personal model. Such technologies can be used to create personalized web services, applications, helpers.

Learn and imagine ...