Transcript Int sys 1 - Intelligent Systems

Architecture of Intelligent Systems - 1
24 Jul 2012
Plan for the Lecture
•I S course – objectives, Arch of Int
systems
•AI Vs Intelligent systems
•Analyse a few systems and get an intuitive
feel for the intelligent systems
•Intuitive understanding of Intelligent
systems
•Understanding the generic characteristics
of the intelligent systems
•Identification of functional subsystems or
blocks
What is an Intelligent System?
It is a concept
It is a System that exists
It learns during its existence
It senses its environment and learns - for each
situation, which action enables it to reach its
objectives
It continually acts, mentally and physically or
externally
By acting it reaches its objectives
It is autonomous in the sense that it decides its
intermediate objectives (to reach its original
objectives) and plans its actions to reach those
objectives, evaluate the consequences of its
actions (by sensing the environment and
judging whether it is a desirable/ expected state)
The actions are deliberate than by chance
It consumes energy and uses it for its internal
processes, and in order to act
Its actions modify the environment
What does this definition imply?
The system has to exist.
•An environment must exist, with which the
system can interact.
•It must be able to receive communications from
the environment, for its understanding of the
present situation.
It must be able to process the sensed information
and build the model of the environment. It must
be able to abstract the communications received
by the senses.
The communications, in turn, need an
interchange of matter or energy.
This communication is for the purpose of
transmitting information
or
for the specific purpose of structuring of
matter in the environment (that the system
perceives).
•The IS must have an objective whether given or
set by itself. Capable of setting its objective
consistent with its original objective.
it must be able to check (or aware) whether its
last action was favorable or has desired impact
on the environment or whether it resulted in
getting nearer to its objective, or not.
•To reach its objective it has to select its response.
A simple way to select a response or an action is
to select one that was favorable in a similar
previous situation. But it may have to reason it
out in new situations.
•It must be able to learn. Since the same
response sometimes is favorable and
sometimes fails, it has to be able to recall in
which situation the response was favorable,
and in which it was not. Therefore it stores
situations, responses, and results.
•Finally, it must be able to act; to accomplish
the selected response.
It is autonomous
Does it mean it has free will
Does it have likings?
Should it know its own capabilities
When should it give up or when
should it pursue an objective
Where do we need Intelligent systems
Intelligent Systems (methods) are used
in science and industry with a wide
variety of applications involving the
following areas: intelligent robotics,
health monitoring applications, speech
and language interfaces, financial
forecasting and prediction, internet and
agent technology, image processing,
planning and scheduling, knowledgebased systems, security and fraud
detection.
AI has given us
• Parsers, theorem provers, inference
engines
• Tools – searching, classification, statistical,
pattern matching, abstraction, translation
• Tools – problem solvers, game playing,
modelling, robotic guidance
• Technologies – neural networks,
knowledge acquisition, expert systems,
planning, dialogue generators
The above are integrated into General
purpose applications
What has it not given us?
How to build intelligent systems by integrating
the above components and techniques.
Reasons
Emulation of Human Intelligence is difficult
Human Intelligence is highly flexible and so no
unique architecture is possible for all
No comprehensive architecture exists as we
have not attempted to solve a problem in its
entirety
We can only simulate one aspect/behaviour at a
time
Intelligence
Architecture
Intelligent systems
Intelligence
Reflex/ known response
Understand, interpret and respond
Perceive, analyse, respond
Respond under unknown and uncertain
environments
Perceive, analyse, generate knowledge (about the
world)(is it learning?), use knowledge and
respond
Plan, compete, Plot, scheme
Evaluate alternatives, judge values?, value system
or beliefs
Set goals for success (who decides the criteria of
success?)
Intelligent systems – Examples
Smart dwellings, Smart space,
Smart vehicles, Intelligent ground, underwater,
space exploratory vehicles
Cave bursters, Humanoid robots, Pilot-less plane
Smart manufacturing plants
Smart security devices – campus security, access
to classrooms and labs – attendance, entry
control, switching off devices, Vehicle/individual
movement an tracking, surveillance
Smart SW to locate other SW
At what stage do we call a system intelligent
Contd.
ITS
Intelligent SW agents
Code generators/translators/ Symbolic
computation
Conversational engine ELIZA
Game engine - Chess engine
Semiotic engines
Intelligent networks
Intelligent systems are
Autonomous – self governing, no external
intervention over extended periods,
Able to work in unpredictable
environment – sense the environment,
decision making, control action
Adaptive
Fault tolerant
Functionality in noisy environment
Autonomy is objective & Control is a means of
achieving it
Contd
Planning
Learning
Re-structurable/re-configurable
systems
Value judgements
Embedded systems
Real time control systems
Control may be in either hardware or
in software
Outline for a theory of Intelligence
IEEE transactions on systems, man, and
cybernetics. Vol 21, No 3, May/June 1991
J S Albus
Intelligent Systems – Architecture, Design
and Control. Albus, Meystel, wiley, 2002
Daniel Dennett.
Brainchildren: Essays on Designing Minds.
MIT Press, 1998.
Intelligence –
Ability to act appropriately in an
uncertain environment
Appropriate action increases probability
of success or decreases the probability of
failure
in the achievement of behavioural goals.
Who generates the criteria for success and
the goals? - Within the IS? or the
environment in which IS operates?
Levels of Intelligence
Sense the environment, take decisions, and
control action
Recognise objects and events, Store and use
knowledge about the world, to reason, & to plan
Perceive and analyse, generate knowledge –
learn, plot and scheme, choose wisely from
alternatives, judge values
Plan successfully in a complex, competitive and
hostile world.
What determines the amount of intelligence
Computational power of the computing
engine
Sophistication and elegance of the
algorithms
Amount and quality of information and
values
Efficiency and reliability of the system
architecture
Amount of natural intelligence can grow through
providing knowledge
learnt through experience
evolved through experience
passed on through generations
Can the Intelligent systems Intelligence grow
through the same approaches i.e. programming,
learning, evolution (modification & genetic)?
Evolutionary Intelligence is a product of
natural selection! Passed on to next
generation, follows natural selection.
In natural selection less successful
behaviour dies out, more successful
behaviour is retained and passed on to the
next generation.
Natural selection is driven by competition of
the individuals in a group & of groups in the
world
How do we implement evolutionary
intelligence in machines? How do they pass
on this intelligence to next generation?
What are the functional blocks of an intelligent
system?
Sensing the world – sensors
Sensor information and value processors
World knowledge generation, storage,
retrieval
Knowledge processors & computational
engine
Inference engine/Reasoning engine &
Decision Support system
Actuators
Control engine