KBAI
Intro to KBAI
Characteristics of AI Problems
- Knowledge arrives incrementally
- Problems exhibit recurring patterns
- Problems have multiple levels of granularity
- Problems are computationall intractable
- The world is dynamic, but knowledge is static
- The world is open-ended, but knowledge is finite
Characteristics of AI Agents
- Limited computing power
- Limited sensors
- Limited attention
- Computational logic is fundamentally deductive
- Agent knowledge is incomplete relative to the world
Knowledge Representations
Problem Solving Techniques
Architectures
Fundamental Processes of KBAI Cognative Systems
- "Cognative" means dealing with human-like intelligence
- Cognative system and KBAI-agent are used interchangably
- Deliberation, Meta-Cognition and Reaction make the three-layer system
Deliberation
Reasoning about the world around us.
- Reasoning
- Learning
- Memory: Knowledge gained from learning needs to be stored
think of being in traffic and wanting to change lanes. Requires planning
Meta-cognition
Reasoning about deliberation or reaction.
Lane-change results in another car honking at you. You need to reason about the cause of the sub-optimal result.
Reaction
- A direct mapping from perception to action
Kinds of KBAI
-
Think/Act
-
Optimally/Like Humans
-
Think like humans: Semantic Web
-
Think optimally: Machine Learning
-
Act like humans: Improvisational Robotics
-
Act optimally: Airplane Auto-pilot
Principles of KBAI
- KBAI agents represent and organize knowledge into knowledge structures to guide and support reasoning
- Learning in KBAI agents is often incremental
- Reasoning in KBAI agents is top-down as well as bottom-up
- KBAI agents match methods to tasks
- KBAI agents use heuristics to find solutions that are good enough, though not necessarily optimal
- KBAI agenbts make use of recurring patterns in the problems they solve
- The architecture of KBAI agents enables reasoning, learning and memory to support and constrain each other
Children