Kognitivnye Systemy is a team of enthusiasts developing a platform for strong artificial intelligence based on neural networks and quantum computing.
The main focus of our work is the creation of interactive systems and smart search systems based on AI.
We are the developers of the first Russian platform for neural networks training - BRAIN2, which is based on our own innovative approach to the creation of AI systems.
With the help of our partners from Promobot, the The Foundation for Assistance to Small Innovative Enterprises, and Skolkovo Foundation we have created an advanced cognitive processor for natural text processing. Today, our models are used in Promobot robots that assist clients throughout Russia and in many countries of the world.
Our goal is to make artificial intelligence technologies as accessible as possible and create the foundation for strong AI in the foreseeable future.

We can help

with the following

tasks

1. Development

of Analytical

Systems
Development of systems based on AI. Our own framework for training models gives us an advantage in speed, size of models, accuracy and flexibility in choosing neural network architectures.
Smart consultants for online stores

- Provide the first level of customer support

- Answer frequently asked questions

- Help customers find the right product

- Learn from real-life dialogues between consultants and customers

Semantic search system for websites and databases

- Training on natural text

- Asking questions in arbitrary form

- Accounting for contextual synonyms

- Correcting spelling errors

Classification of applications and appeals

- Classification of appeals by topic

- Automated selection of a suitable response template

- Highlighting key named entities - names, locations, dates

- Segmented case analytics

Conversation module for promo-robots

- Communicating with people on a wide range of topics

- Providing consultations on products or services

- Ability to define conversation goals

- Consideration of synonyms and paraphrases

Smart consultants for online stores

- Provide the first level of customer support

- Answer frequently asked questions

- Help customers find the right product

- Learn from real-life dialogues between consultants and customers

Semantic search system for websites and databases

- Training on natural text

- Asking questions in arbitrary form

- Accounting for contextual synonyms

- Correcting spelling errors

Classification of applications and appeals

- Classification of appeals by topic

- Automated selection of a suitable response template

- Highlighting key named entities - names, locations, dates

- Segmented case analytics

Conversation module for promo-robots

- Communicating with people on a wide range of topics

- Providing consultations on products or services

- Ability to define conversation goals

- Consideration of synonyms and paraphrases

Markup of natural text data for the purpose of its use in training language models. Thanks to our own NLP processor and the experience of our specialists in working with various types of text data, we can offer a unique complex of technical and analytical markup of natural text.

3. Testing Scientific

or Business Hypotheses

We help to collect and mark up data, train models and test your hypotheses. Our Data Science specialists are skilled in a variety of methods for working with data and various types of neural networks.

Market Ready

Solutions

BRAIN2 is the first Russian neural network training platform, based on our own framework.

BRAIN2 allows developers to create predictive models on their own data, test the effectiveness of models, and then use them to make predictions. The platform greatly simplifies the process of neural network training by automatically selecting the optimal model architecture.

You can use our online platform with access through your personal account or purchase a license to install the system on your machine.

BRAIN2NLP - is a Russian startup which develops artificial intelligence systems based on neural networks.

Main features of BRAIN2NLP include language detection, stemming, morphological and syntactic analysis, named entity recognition and determining the semantic vector of words.

This allows models trained with BRAIN2NLP to better understand the context and provide more relevant answers to user’s questions.

Project

“Consciousness”

We have developed a theoretical basis for creating a complex system with an energy function and the ability to independently generate algorithms of its actions.

This will allow the creation of artificial intelligence, the main purpose of which will be to maintain its vital functions and compete with other models for survival. The capabilities of such artificial intelligence will significantly exceed the technologies used today.
Our Advantages
A unique patented approach to create AI models (speed, simplicity, model’s size)
Over 50 completed projects
30 years of team experience in AI systems development
Kognitivnye Systemy team participate in the NTI working group, Neuronet, are the authors of inventions and patents in the field of AI
Kognitivnye Systemy is a Skolkovo resident and winner of grant competitions from the Foundation for Assistance to Small Innovative Enterprises
Team members regularly publish scientific articles

Our

Projects
MFC service

We have developed software for Promobot robots which "work" in government multifunctional centers.

It allows the visitor to obtain MFC service information upon the user’s request made in arbitrary form.

MFC service

We have developed software for Promobot robots which "work" in government multifunctional centers.

It allows the visitor to obtain MFC service information upon the user’s request made in arbitrary form.

Smart search engine brain2semantic search

It is a robot assistant that “reads” the text of any volume and subject in a small amount of time and then answers the questions in natural language. The service is based on our own Semantic Vector Model, which is aimed at finding semantic links in the text and has the ability to take into account the context for each word.

Smart search engine brain2semantic search

It is a robot assistant that “reads” the text of any volume and subject in a small amount of time and then answers the questions in natural language. The service is based on our own Semantic Vector Model, which is aimed at finding semantic links in the text and has the ability to take into account the context for each word.

Museum Guide

A question answering system for a robot, which allows it to answer visitors' questions in a arbitrary form and inform about interesting historical facts. Today the robot is already conducting tours for museum visitors.

Museum Guide

A question answering system for a robot, which allows it to answer visitors' questions in a arbitrary form and inform about interesting historical facts. Today the robot is already conducting tours for museum visitors.

Scientific articles

and

Media

The art of the possible

“The key concept here is meaning,” notes Artem Artemov. – The relationship between the cause (for example, the red light of a traffic light) and the effect (stopping the machines). The second component of “humanoid” intelligence is feelings …

We need them to create motivation and find meanings on our own. And who will appreciate these meanings if the AI ​​cannot tell anyone about them? Therefore, we need a third element – the ability to communicate effectively ”

https://mipt.ru/upload/iblock/e65/2017_zanauku_1.pdf, pages 58-61
Valentin Malykh, Ksenia Ulanova

Meet your new personal assistant. What is artificial intelligence capable of?

“We are sure that it is impossible to create artificial intelligence without teaching it to feel. We believe that artificial intelligence is only able to know when something will move it …

After all, our life is the path from suffering to pleasure. It would be nice to drive artificial intelligence there ”- Artem Artemov, Cognitive systems

https://rtvi.com/stories/znakomtes-vasha-novaya-personalnaya-pomoshchnitsa-na-chto-sposoben-iskusstvennyy-intellekt/
Sergey Morozov

Artificial intelligence in retail is needed to optimize performance.

Artem Artemov, CEO of Cognitive Systems, believes retailers should not have service personnel left in the near future, intelligent assistants will answer frequently asked questions.

“For the former analyst, part of the work will be done by the intellect, and the freed up time can be spent on something else. Do not be afraid that people will be left without work. Those who are afraid of changing with a changing world will be left without work. ” – claims Artem Artemov.

Искусственный интеллект в ритейле нужен для оптимизации работы

Waiting for miracles, ordinary miracles.

A. Artemov, A. Sergeev, I. Khasenevich

Intelligent computer programs come on the heels of representatives not only simple, but even creative professions such an interpreter and journalist, threatening in the near future to oust them from the market.

According to the UN report, robots will soon take 2/3 of the available jobs in developing countries. Let’s try to understand how justified science fiction films and robot-centric forecasts are, and whether it is possible to talk about the development of real artificial intelligence.

http://strf.ru/material.aspx?CatalogId=222&d_no=126692#.XCXck88zacZ

Nonrandom number generator and meaning understanding problems

Report of the Kognitivnie Systemi CEO A.A. Artemov at OpenTalks.ai conference ” Nonrandom number generator and meaning understanding problems “. 2019

Accounting for unknown features in the model on the example of the MFC case

Report of the Kognitivnie Systemi CEO A.A. Artemov at Big Data & AI Conference 2019 ” Accounting for unknown features in the model on the example of the MFC case”. 2019

A Method for Estimating the Proximity of Vector Representation Groups in Multidimensional Space. On the Example of the Paraphrase Task.

Artemov, B. Alekseev. The following paper presents a method of comparing two sets of vectors.

The method can be applied in all tasks, where it is necessary to measure the closeness of two objects presented as sets of vectors.

https://arxiv.org/abs/1908.09341

A Method for Estimating the Proximity of Vector Representation Groups in Multidimensional Space. On the Example of the Paraphrase Task

A. Artemov, B. Alekseev. The following paper presents a method of comparing two sets of vectors.

The method can be applied in all tasks, where it is necessary to measure the closeness of two objects presented as sets of vectors.

https://arxiv.org/abs/1908.09341

The Training of Neuromodels for Machine Comprehension of Text. Brain2Text Algorithm

A. Artemov, A. Sergeev, I. Khasenevich, A. Uzhakov, M. Chugunov

Nowadays, the Internet represents a vast informational space, growing exponentially and the problem of search for relevant data becomes essential as never before. The algorithm proposed in the article allows to perform natural language queries on content of the document and get comprehensive meaningful answers.

The problem is partially solved for English as SQuAD contains enough data to learn on, but there is no such dataset in Russian, so the methods used by scientists now are not applicable to Russian.

Brain2 framework allows to cope with the problem – it stands out for its ability to be applied on small datasets and does not require impressive computing power. The algorithm is illustrated on Sberbank of Russia Strategy’s text and assumes the use of a neuromodel consisting of 65 mln synapses. The trained model is able to construct word-by-word answers to questions based on a given text. The existing limitations are its current inability to identify synonyms, pronoun relations and allegories. Nevertheless, the results of conducted experiments showed high capacity and generalisation ability of the suggested approach.

https://arxiv.org/abs/1804.00551

Informational Neurobayesian Approach to Neural Networks Training. Opportunities and Prospects

Artem Artemov, Eugeny Lutsenko, Edward Ayunts, Ivan Bolokhov

A study of the classification problem in context of information theory is presented in the paper. Current research in that field is focused on optimisation and bayesian approach.

Although that gives satisfying results, they require a vast amount of data and computations to train on.

Authors propose a new concept named Informational Neurobayesian Approach (INA), which allows to solve the same problems, but requires significantly less training data as well as computational power. Experiments were conducted to compare its performance with the traditional one and the results showed that capacity of the INA is quite promising.

https://arxiv.org/abs/1710.07264

Contact us
Your name
Email
Phone