Transcript powerpoint
Artificial Intelligence and Lisp #2
Introduction to Cognitive Agents and to
Knowledge Representation
Software agents
A piece of software that produces a behavior in
terms of discrete 'actions', and which is
perceived as an entity 'doing' these actions
It is the instance of the software that is an agent
Autonomous agent: decides itself
Model-based agent: uses a model of its
environment for selecting actions
Cognitive agent: a model-based agent using a
concept-based model of the environment
Uses of concept-based model
Know procedures/ scripts/ methods
… and be able to apply them
Diagnose problems and resolve them
Imagine what will happen
Use earlier experience and adapt it (learning)
Have facts and apply them
Acquire facts
Additional uses of such a model
Adapt earlier solutions to problems
Identify relevant facts
Structure a given problem and its solution
Draw conclusions from selected facts
Identify and apply constraints
Concept-based model supports
Robustness
Flexibility
User-friendlyness
Scenario for this course: the Zoo
Entities: animals, spaces for animals, food for
them; zookeeper, guardians, veterinary, ...
Actions and events: move an animal, feed an
animal, animal gets ill, treat illness, animal gives
birth, animal dies, …
Each course participant 'builds' his or her zoo
and applies his/her agent to it
Then (maybe) we connect the zoos together
Scenario for this lecture: Household
Live illustrations
These lead to the introduction of a notation
which is presented in the lecture notes (part I)
Agent behavior frameworks
Command-taking agent, with a certain
intelligence in carrying out the commands
Monitoring agent, in charge of maintaining the
correct state in its environment
Plan-executing agent, in charge of performing a
given plan and monitoring that it proceeds as
intended
More complex cases?
Preparation framework
Two phases: preparation phase and
performance phase
Example: prepare for a reception
Performance phase: guests arrive and are
entertained
Preparation phase: think through what must be
done in order to facilitate the performance
phase.
Preparation framework (revised)
Two phases: preparation phase and
performance phase
Example: prepare for a reception
Performance phase: guests arrive and are
entertained
Preparation phase: identify situations that may
arise, actions that will be required for dealing
with these, and preconditions for those actions.
Arrange that preconditions are satisfied.
Software architecture for AI applications
Preparation
Monitoring
Command
execution
...
Planning
Diagnosis
Dealing w
obstacles
...
Model-based agent - MIA
Cognitive system platform - Leonardo
Programming language - CommonLisp
Operating system
Is there an architecture for
“universal artificial intelligence” ?
SOAR proposal (Laird, Newell, Rosenbloom)
presumes five steps performed cyclically:
Input
Elaboration
Decision
Application
Output
… and with possibility of recursion