Download Some Thoughts to Consider 1

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Some Thoughts to Consider 1
•
What is so ‘artificial’ about Artificial Intelligence?
•
Just what are ‘Knowledge Based Systems’
anyway?
•
Why would we ever want to study this stuff?
•
Mind is what the brain does.
•
‘Brains cause minds’ - Searle.
•
Software is to machine as mind is to brain.
•
‘Knowledge is power’ - Francis Bacon.
•
Can machines think?
•
What does it mean to build software systems
that are ‘people-literate’, rather than having
people be ‘computer-literate’?
•
Does the study and use of AI help us better
understand how people think and act?
Anticipated Benefits of Investing in
Emerging Technologies
•
Particularly the technologies of:
•
•
•
•
•
Knowledge based systems
Agent oriented systems
Service oriented architectures
Neural networks
Genetic algorithms
•
Move people to a new level of problem solving.
•
Raise business concepts and operations to a
higher level of understanding.
•
Manage the increased complexity of running the
business.
•
Reduce the time required to field new
applications.
•
Produce more intelligent performance
enhancement applications.
•
Reduce long term system maintenance time.
•
Provide bottom-line value to clients and profit for
the corporation.
The Main Design Issues
•
•
•
Representation
•
What structures or ‘metaphors’
shall be used?
Knowledge
•
Where and how shall it be
represented?
Process Control Flow
•
Where in the architecture shall it
reside?
Types of Knowledge
•
Facts
•
Process Knowledge
•
Operational Know-How
•
Market Knowledge
•
Technology/System/Database Knowledge
•
Dependency Knowledge
•
Causality Knowledge
•
Conflict Knowledge
•
Constraint Knowledge
Types of Knowledge
•
•
Concept Knowledge (Objects, Nodes)
•
•
•
•
Physical objects
Actions
Events
Categories
Relationship Knowledge (Links, Arcs)
•
•
•
•
•
A-kind-of
Part-of
Instance-of
Cause-of
Acts-on
•
Descriptive Knowledge (Attributes)
•
Procedural Knowledge (Algorithms)
•
Inheritance Knowledge (Classes)
•
Heuristic Knowledge (Rules of Thumb)
•
Inference Knowledge (Strategies)
•
Emergent Knowledge (Neural Nets)
Types of Representation
•
Declarative (Facts)
•
Procedural (Instructions)
•
Inferential (Implied by Reasoning)
Mechanisms of Representation
•
•
•
•
•
•
•
•
Rules
Frames
Predicate Logic
Semantic Networks
Classes – Objects – Methods
Actors – Agents
Neural Nets
Genetic Algorithms
Key Knowledge Engineering Activities
•
•
•
Knowledge Acquisition
•
•
•
•
Interviewing experts
Protocol analysis
Prototype iteration
System acquisition of knowledge (learning)
Knowledge Representation
•
•
•
Categories of the knowledge
Structure of the knowledge
Tool selection
Knowledge Utilization
•
•
•
•
•
Control structure – “knowledge flow”
Reasoning strategies
Justification and explanation
Dealing with uncertainty and incompleteness
System validation
So, What About Decision Support?
•
We are evolving a new kind of product.
•
•
One that is knowledge-enriched, with locally-authored
decision support.
Rather than a vendor-supplied, predetermined package
of software logic and data structures.
•
This requires intense knowledge engineering
and knowledge representation that is
substantially different from traditional
programming practice.
•
Knowledge is represented declaratively in a
knowledge base such that customers can
customize it for local use.
•
Knowledge is not represented in programming language
code.
Model Based Software Design
•
Represents a different way of thinking about
software design and implementation.
•
Takes the clinical (business) knowledge out of
the Java code.
•
Moves the problem solving process to a higher
level of abstraction.
•
Models become the vernacular for clinical
(business) architecture discussions.
•
Representation is ‘outside’ the Java classes,
rather than ‘inside’ the Java classes.
•
The Java classes become more like ‘engines’
that manage and reason over the external
representations.
•
The movement to XML, RDF, and OWL is
movement in this design direction.
Motivation for Model Based Architecture
•
We’re growing out of traditional ‘database-toscreen’ types of product.
•
We are faced with providing more
‘knowledge-rich’ products.
•
Customers require customization of the
content we deliver for their specific product
venue.
•
More and more of our traditional products
require integration and interoperability.
•
Analysts are required to participate more in
the design of representational structures.
•
Developers are required to participate more in
product design.
•
The level of complexity of our products is
increasing beyond what is manageable by
traditional development means.