Biologically motivated cognitive architectures (BICA) - free course from Open Education, training 10 weeks, from 2 to 3 hours per week, Date November 28, 2023.
Miscellaneous / / November 30, 2023
This course is offered to master's students. BICA is a promising, rapidly developing field at the intersection of artificial intelligence, biology and cognitive science. One evidence of this is the growing number of scientific publications related in one way or another to BICA. Here, cognitive architecture is understood in a broad sense, as a template for the development of intelligent agents. The sources of biological motivation are the brain (neuroscience) and human thought (cognitive psychology). The course will ensure that students develop basic knowledge in the field of cognitive architectures, their basic elements and principles, approaches to their implementation, their study and use in virtual environments. Students will learn about global artificial intelligence problems and BICA-based approaches to solving them, as well as the tests and metrics used for assessment. Some of the key concepts and topics that underlie BICA will be covered in detail, including human memory systems, neural network models, semantic mapping, common sense reasoning, etc. Particular emphasis will be placed on the roadmap to solving the BICA Challenge and promising applications of future BICAs humanoid type.
The course is bilingual. The material is presented mainly in English with Russian subtitles.
Module 1. General introduction.
Can a machine have a consciousness similar to that of a human? Ambitions and problems of artificial intelligence (AI). Cognitive architectures as an alternative approach to creating AI. Interest in this area in the scientific world. Cognitive architecture research communities.
Basic information from cognitive psychology: introspection, behaviorism, cognitive revolution and computer analogy of the brain.
Models of human memory systems, explicit and implicit, short-term and long-term memory. Elements of the cognitive cycle, perception, attention, imagination.
Module 2. Introduction to Neuroscience.
A brief introduction to neuroscience: elements of neurophysiology and neuroanatomy, behavioral, computational, systems neuroscience. Psychophysiology, brain imaging, and cognitive neuroscience.
Principles of operation of neurons and their elements. Behavioral correlates of neural activity. Types of coding. Localization of functions. Examples: stimulus detectors, mirror neurons, place cells, granny neurons. Binding problem. Discussion about the nature of imagination.
Module 3. Biological and machine learning of neural networks.
Mechanisms of memory formation in the brain. Neural network models and attractors, their types and connections with biology and psychology. Spatial cognitive maps in biology. Their role in memory formation.
Elements of the theory and applications of neural networks. Evolutionary programming and other forms of machine learning. Possibility of connection with biology.
Module 4. Knowledge representations and semantic mapping.
Concepts of sign, symbol, language. Representations of concepts and categories in human memory. Semantic networks and connectionism. Semantic lattices and concept analysis.
Continuous semantic spaces. Strong and weak semantic maps. Semantic mapping methods: mathematical, physiological, psychological, and linguistic aspects. Types of semantic maps and their applications. Semantic mapping of brain activity and “mind reading.”
Module 5. Principles, diversity and evolution of cognitive architectures.
Evolution of approaches to creating intelligent agents. The concept of cognitive architecture. Cognitive architecture as an embodied intelligent agent, as a programming language, and as a theoretical framework.
General theory of cognitive architectures. Memory systems, cognitive cycle. Hierarchy of cognitive architectures. Trends in the expansion and merging of BICA models. Common minimal cognitive model (Common Model of Cognition) and the most extended functional diagram of BICA. The concept of critical mass.
Operating principles of the most famous specific cognitive architectures: Soar, Act-R, Clarion, Icarus. Hybrid BIKA. Overview of the diversity of BICA models. GMU BICA example. Table of cognitive architectures.
Module 6. Emotion modeling and emotional cognitive architectures.
Types of computational approaches to emotion modeling. Discrete and component models. Affective spaces. Logical and statistical approaches: modal logics, situation calculus, BDI models, inductive inference methods. Examples of emotional cognitive architectures (EMA).
Why does a robot need a sense of humor? The problem of modeling complex and social emotions. Moral schemes. eBICA example.
Module 7. Memory of the past and the future, the possible and the impossible.
Episodic memory. Prospective and retrospective autobiographical memory. Consolidation and reconsolidation. Retrograde and anterograde amnesia. "Theory of Thought". Concepts of “I”, memory manipulation. Free will, determinism, trust.
Types of metathinking. Social and narrative intelligence. Fable and plot. Character and role. Author and actor. Narrative network and work scenario. Narrative planning, autonomous goal generation, believable characters. Socially acceptable intelligent agents.
Module 8. Human learning, BICA and the path to AI critical mass.
The problem of teaching in pedagogy. Types of training. Active learning. Learning through reasoning and problem solving. Self-regulated learning. Meta-learning. The role of emotions, imagination, social and metathinking in the realization of learning ability.
Implementation of theories and models of human learning in a computer. Intelligent tutoring systems based on BIKA and their application in the educational process. The task of creating a general purpose “artificial student”. Overcoming the barrier in human consciousness.
Module 9. Applications of cognitive architectures.
Scientific and practical problems solved on the basis of BIKA. Applications in medicine, psychology, military affairs, social engineering and analytics, education, business, art, entertainment, etc. Artificial creativity.
Module 10. Systems and methods for assessing cognitive architectures and AI development.
Tests, criteria and metrics for assessing the performance of intelligent systems. Cognitive Decathlon. The Turing test and its modifications. Virtual environments and VR environments for studying the behavior of natural and artificial cognitive architectures during their social interaction. Efficacy, credibility, and social compatibility. Intellectual and social-emotional competence. Application of characteristics of the human psyche to artificial systems.
Setting the task of creating strong AI. Possible options for the development of AI. Possible role of cognitive architectures in AI systems of the near future. Challenges, dangers and road maps. Ethical and philosophical issues.