Creating a system of strong artificial intelligence with self-awareness and knowledge about the world. The intelligent system has to be able to understand the outside information. Our concept suggests such understanding to derive from building a set of possible models of the world. These models can explain the past and predict the future. Each model has a degree of credibility, which allows them to be ranked, choosing the most likely ones for analysis.
Tasks
• Realization of understanding of incomplete and ambiguous verbal information.
• Communication with the system in an unambiguous formal language and gradual transition to a natural language.
• In-depth description of everyday knowledge for a limited set of meanings.
• Testing the methods for learning and automatic knowledge change for an intelligent system.
• Self-modeling by introducing the dedicated "I"-object into the world model.
Methods
Methods
• Fuzzy and probabilistic logic
• Semantic networks
• LSTM and Word2Vec deep neural networks
Tools
• DemonScript knowledge and logic description language
• C++, Python, JavaScript
Team
A team of 15 is working on the project, including developers, knowledge engineers and linguists.
DemonScript
DemonScript is a language designed to present everyday knowledge and draw logical conclusions in the context of incomplete information. The basic construct with which DemonScript works is a semantic network consisting of nodes connected by directed edges.
For more complete documentation DemonScript (Russian).