Core Expert is a complicated software system, including various methods of storing and processing knowledge. It is a hybrid model consisting of many layers of knowledge at a different level of representation and abstraction. They can be configured in such a way that on the basis of the core it is possible to assemble intelligent systems for various purposes: from intelligent assistants to control systems in robotics. Continuous development and improvement of the core will allow it to be used in any sphere of human activity.
The concept of hybrid models in artificial intelligence
There are many approaches and methods in artificial intelligence.
These are formal grammars, genetic algorithms, semantic networks, neural networks, frames, fuzzy algorithms, and others.
Each of the existing approaches solves a certain range of problems, has its advantages and disadvantages, limiting its application.
But any real task, such as driving a car, natural dialogue, enterprise management in a broad sense, training, translation and others are
The AI-complete task is a task that can be solved by an artificial general intelligence (AGI) that has a level of intelligence no lower than the human and able to make decisions under uncertainty, plan, train and combine all these properties to achieve common goals.
To solve complex problems, two or more approaches are used together.
Let's call a
weak hybrid a model that includes several basic approaches, interconnected by some interface.
An example of such a model can be the determination of the basic parameters of a neural network using genetic algorithms.
strong hybrid will be called a model in which the technologies are so strongly embedded in each other that in fact there is a new technology, ideally taking the strengths of each of the models.
An example is fuzzy neural networks that combine a fuzzy logic apparatus and neural networks.
A core is a collection of components, which we call
Each layer combines knowledge and methods of processing them at some level of abstraction.
The layer can be created within the framework of some known model, but can also be a strong hybrid.
The development of some strong hybrids for ten years has become the basis for the creation of a universal intellectual core.
These hybrids (selected technologies, method of integration, form of algorithmic representation) are, at the moment, intellectual and commercial secret.
We plan to disclose some information about them in publications, after checking the practice and creating certain products.
Layers are organized in a hierarchical structure in such a way that for different specific products it is not necessary to use all of them - just select the necessary ones. Between each other, the layers are linked by the principle of a weak hybrid. The interface between the layers is another key core technology. In addition, the layers can be of two types: physical and logical. The physical layer represents a specific technology, the logical layer is a specific set of knowledge.
Some layers of the core
Abstract - ideal
Abstract knowledge about the world - its structure and basic relationships
Knowledge of the world in all the diversity of facts
Knowledge about how to work with knowledge
The personality model for an intelligent assistant
Knowledge of the dialogue, its strategies and elements
Knowledge of images and their recognition
Knowledge of sounds and speech model
Multi-level memory model
Want to see more?
The work is organized like a spiral, at each new stage we add new opportunities and return to the existing layers and bring them to a new level.
Work inside the core is performed by a hierarchical multi-agent system. The implementation of the core is divided into stages. At the moment we finished the fourth stage.
Now we are at the start of a fifth stage. Stay in touch.