It is time to open your eyes!

And learn that Neural Networks and Artificial Intelligence are not the same statements.

Existing approaches and technologies in Artificial Intelligence (by Roman Dushkin)

AGI technology by MSL and its general structure

Our technology is based on the Intellectual Core. The Intellectual Core is a complex software package that includes various methods for storing and processing knowledge. Core is a hybrid model that consists of many layers of knowledge at different levels of representation and abstraction. Due to the modular structure, these layers can be arranged and configured so it is possible to assemble intelligent systems for various purposes on the basis of the core: from intelligent assistants to robot control systems.

Each layer combines knowledge and methods of its processing at some level of abstraction. A layer can be created within the framework of some kind of model. Also it can be a strong hybrid that uses several approaches to solve problems at once.

The layers are organized in a hierarchical structure in such way that it is unnecessary to use all of them every time — just to select the necessary ones.

The layers are interconnected by the principle of a weak hybrid. Moreover, the layers can be divided into two types: physical and logical. The physical layer represents a separate technology, the logical — a separate body of knowledge.

The work inside the core is performed by a hierarchical multi-agent system.

This approach allows us to find new practical ways of development and go beyond the limitations that researchers in the field of Artificial Intelligence face.

What makes us different

Most modern approaches to AI rely on the Deep Learning. This approach has been successful with recognition and synthesis of speech, as well as in problems of clustering and decision making. But this approach, in its current and predicted form, is not very suitable for creating general artificial intelligence and has many drawbacks.

In our laboratory we use the descending approach, modeling the higher intelligence functions first. We don't use one specific task – we create a technology for solving any problem within the framework of a system. We use at once several approaches to present and work with knowledge and take into consideration the future interaction of several AI cores in a single ecosystem.

Some of Techs We Use

  • Semantic Networks
  • Finite-state Machine
  • Petri Nets
  • Production Systems
  • Fuzzy Logic and Fuzzy Algorithms
  • Frames
  • Multi-agent Systems
  • and other unique developments

Key Differences

  • Hybrid Model of Knowledge Representation
  • Figurative thinking
  • Abstract Model of the World

  • Multi-layer Memory
  • Self-aware Thinking Model

The most important advantages of the Hybrid AI technology by MSL

Explainable AI

All AI’s chains of thought and lines reasoning can be tracked, documented and explained

Easy Transfer Learning

The structure of knowledge of AI allows to transfer it from one core to another without any leakages of data. Moreover, it is possible to input knowledge in the system as a plain text.

One Iteration Learning

The system can be trained with any amount of raw data. All training requires one iteration. Also there are methods to control and edit new knowledge.

Structured Prediction and Learning

Knowledge is arranged in a multi-level hierarchical structure. Learning mechanisms allow to check the consistency of new and previously obtained knowledge.

Free of Catastrophic Forgetting

AI’s knowledge can be extended without losing the previously obtained knowledge. In addition, there are mechanisms of forgetting which do not lead to the loss of knowledge integrity.

Incremental Learning

Knowledge can be gradually accumulated in the system. It is possible to store contradictions, vague knowledge, etc.

Lauguage Independent Figurative Thinking

The core processes and stores information in a form independent of the natural language.

Module Structure

It is possible to assemble a solution suitable for certain tasks from the components of the core. This solution will use all the power of the core and will not include models that are unnecessary for the task.

Offline Operation

The core can operate isolated from the Internet and cloud servers.

Any “Weak Hardware”, any OS

The structure of knowledge and work algorithms is developed without any dependencies on the operating system or hardware. All you have to do is just to transfer the core of the AI from one operating system to another.

Also, there are no combinatorially complex calculations in the work of agents, which allows you to run the core on a "weak hardware".

Multi-layer Memory

This is a feature of architecture that allows you to store knowledge in various sections. It gives the flexibility to represent any objects and phenomena in the intellectual core.


Each core layer is optimized for certain types of knowledge and stores them in the most compact form. It allows to significantly save resources.

At the moment, the core of AI and all its knowledge take about 7 megabytes.

Check out the presentation for more information

    Frequently Asked Questions


    • What type of neural network do you use?

      We do not use neural networks to model higher intellectual functions. At the moment, they are completely absent in the core technology.

      Our approach is called the "Hybrid Model of Knowledge Representation" and takes advantage of various approaches in the implementation of AI, as well as our own unique laboratory developments.

    • Siri and Google Assistant cannot do what you declare. Do you have more training data than they have?

      The work of such systems is based either on scenarios that developers manually write or on the basis of neural networks that train on user dialogs. They work with data, not knowledge.

      Features of the knowledge representation of our system allow us to create a picture of the world in a different form: abstract, factual, logical, in the form of tasks.

      Each type of knowledge uses its own specific approaches and algorithms. It allows to create a complex but flexible structure with many features that do not require huge amounts of raw data.

    • How do you solve the problem of the "Chinese room"?

      The work of our technology is not based on working with language symbols, but on processing abstracts. Therefore, the AI understands and realizes that it knows how, with what data it works, how it makes decisions, and so on.

      For example, when the word "dog" occurs, the AI works with all its abstract at once, and not with the word.

    • How do you train your AI?

      We have a set of special modules to simplify and accelerate the training of certain types of knowledge.

      Learning takes place in an iterative approach. We periodically return to what has already been studied to expand and refine knowledge. Separate knowledge itself is accepted and remembered by Artificial Intelligence in one iteration and does not require a lot of iterations.

    • How long do you work on this?

      The first research and design of the system began more than 10 years ago. The last 2.5 years we have been working on implementation in practice.

    • What is your main goal?

      Mind Simulation is a research laboratory whose mission is to create General Artificial Intelligence.

      Our main goal is to achieve General Artificial Intelligence and maximize its application in various fields.

    • What problem do you solve?

      The solving of the problem of creating an AGI and even the first steps to it open up many additional opportunities in solving complex intellectual problems.

      A full-fledged intelligent assistant, a “live” NPC in the game, full-fledged driving: all this is impossible without AGI technology.

    • Who is Steve and why he named that?

      Steve is the main Artificial Intelligence, which the laboratory is gradually developing. He is much more intelligent and advanced than, for example, Geralt or everything else that we show.

      We work exactly with Steve when we develop technology and move it forward, adding new modules, knowledge and capabilities that are embedded in our roadmap. Gradually, as other areas grow, we transfer some opportunities from Steve to projects, such as Geralt.

      Steve will be the core of a future project that we have not yet announced.

      Why is his name so? World history knows many outstanding personalities. When we thought about what name to give the main AI, we suddenly realized that it absorbs the best qualities of some people, whose name is Steve.

      This is Steve Jobs – a person who could understand what we want and, most importantly, explain it to us, a quality leader of the company; Steve Wozniak – a brilliant engineer; Steven Spielberg – a man with great imagination; Stephen Hawking – one of the greatest minds of mankind and Stephen Strange – a fictional character with the highest level of intelligence.

    • Where can we test Steve?

      Right now we are not showing Steve publicly. His time is yet to come.

    • You have B2B Department. What is it?

      B2B Department is engaged in the development of the Business Intelligence product, whose task is to optimize business processes within the company and automate external and internal communications.

      At the moment, we are in an advanced stage of a pilot project with a large medical holding and will be ready to present the product in the coming months.


    • What is it?

      This is a toolkit that makes it possible to endow all game characters (NPCs) with mind using Artificial Intelligence. That is, to create a digital personality that is able to communicate and behave like a real person outside of the prepared replicas.

    • Who is this product for?

      Technically, this product is created for developers of video games. But the most important are gamers here, because for them the game developers will create a new gaming experience.

    • When will you release it?

      At the moment, we are focused on demonstrating the capabilities of the technology and testing it in the “field” by revitalizing Geralt from Rivia and some of his friends.

      The completion of this journey or a pilot project with a game studio will give us the opportunity to name the dates.

    • Why voice is so robotic?

      At the moment, voice is not a priority. We have shown that this is possible and works offline. Intellectual capabilities, advanced personality settings and knowledge of the world are now in priority. The most important thing is that the AI thinks and understands the person, and the voice is only one way of interaction.

      Recently, we wrote that we confirmed our concerns and were very disappointed in the current state of voice technologies. They cannot reveal all the capabilities of the AI core and severely limit them.

      Therefore, we will immediately have to completely make our solution. We already have a work plan and a technology concept. We will begin implementation when we are ready to allocate human resources to solve this problem.

    • How gamedevs will be able to tell stories if all NPC will be alive?

      It is important to understand that live NPCs in games can be used in different ways. In the first case, these are fully independent units, and in the second - controlled ones. Let's take a closer look:

      1. Here we can give an example of an online game, a sandbox, where all power is given to the players. Such NPCs may be merchants; residents of cities, planets; companions and so on.

        Actually, that also applies to "story" games. For example, The Witcher 3, there are many NPCs that are not involved in the story. Now they are just decoration, but they could give much more interactivity.
      2. These are important NPCs that are actively involved in the storytelling. They can also be independent, but developers will be able to lay in them an irresistible desire to do or say something at a certain moment so that the story they want to tell evolves.

        Remember how GladOS told Chel why she runs tests all the time? She said that she began to “itch” if she was not doing research. There is something in common here.


    • Do you work 12 years on Geralt? How will game devs create thousands of NPC's personalities?

      No. We have not been working on Geralt for 12 years. We have been working on the Witcher Universe for about 8 months and this work has nothing to do with working on the core technology.

      For 12 years, we have been engaged in research and development of Artificial Intelligence, and Geralt “enjoys” the fruits of the work already done, which is imposed on knowledge of the Witcher Universe.

      Working with knowledge within the framework of the Witcher Universe is only once. After all, the idea of the world for all characters is the same. Only the volume of knowledge (outlook), personality and their features differ. This means that after working knowledge once, we can quickly create thousands of different characters.

      The main characters can be worked out in more detail, but even these terms are not put in any comparison with the time that developers spend creating the game.

    Do you still have questions left?

    Artificial General Intelligence laboratory
    © Powered by Mind Simulation.