80 likes | 256 Views
CPSC 327. Week 2 AI as Representation and Search. Knowledge Representation. Most important part of AI How do we get objects and relationships from the world into the computer. Suppose we have a stack of blocks: A B C And this goal C B A And these rules:
E N D
CPSC 327 Week 2 AI as Representation and Search
Knowledge Representation • Most important part of AI • How do we get objects and relationships from the world into the computer
Suppose we have a stack of blocks: A B C And this goal C B A And these rules: • If x is clear, put x on the table • If x is clear and y is clear, put x on y
Notice that we say nothing about color, texture, weight. A KR is an abstraction of the essence necessary for solution from objects that exist in the world So, we have only • Name of the block • Position of the block relative to other blocks
Given the initial state and repeated application of rules, we can generate this graph: A B C R2 R1 B C A R2 R1 R2 C B A A B C B A C R2 R1 A B C
Called a State Space Could • Assign a metric to the goodness of each state, depending on how close it is to the goal • Representing each state with a 3X3 array • Notice • Representation is a surrogate for the real world • Representation is an approximation. The only representation that’s error-free is the thing itself • Representation selects important properties • Representation sanctions two inferences • Blocks are/are not in goal state • Some states are closer to goal than others
Many Representation Schemes in AI • First order predicate logic • Semantic nets (structured inheritance networks) • Frames and Scripts (Stereotypes) We’ll look at each.
Important Point AI is concerned with: • Qualitative rather than quantitative problem-solving • Reasoning rather than calculation • Organizing large amounts of varied knowledge rather than implementing a single well-defined algorithm • Good-enough rather than optimal solutions • Means that representation must have structures that are more complex than traditional representational formalisms like arrays.