Introduction Dr. Ahmed Rafea CSCI485 Intelligent Agents 1 Chapter - - PowerPoint PPT Presentation

CSCI485 Intelligent Agents 1

Introduction

• Dr. Ahmed Rafea

CSCI485 Intelligent Agents 2

Chapter Outline

Artificial intelligence Internet and the Web Intelligent Agents Events, Conditions and Actions Agents and Human-Computer Interfaces Using Java for Intelligent Agents

CSCI485 Intelligent Agents 3

Artificial Intelligence

About 40 years old What was hard for people and easy for

computers was minimal compared to the things that were easy for people but almost impossible for computers

CSCI485 Intelligent Agents 4

Basic Concept

early idea included game playing and math after 40

years, there are 3 major phases of development:

1) early years: formal problems, emphasis was on creating

general thinking machines

2) most successful AI projects aimed at very narrow problem

domains and usually encoded much specific knowledge about the problem. Led to Expert Systems.

3) working on problems of human capabilities. The explosive

growth in the internet and distributed computing has led to the idea of agents moving through the network and interacting with each other.

What do we mean by intelligence? An intelligent agent

acts rationally, it does the things we would do, but not necessarily the same way.

CSCI485 Intelligent Agents 5

Symbol Processing

• Intelligence is
• being able to recognize situations or cases
• being able to learn from examples and generalize and apply that knowledge

to new situations

• Produce intelligent behaviour by manipulating symbols
• A problem must be represented by a collection of symbols
• Then appropriate algorithm developed to process symbols
• Symbols are tokens representing real world objects or ideas
• Ways to manipulate symbols:
• use in if-then rules, processed using reasoning techniques called forward

and backward chaining

• construct a semantic network where the symbols and the concepts they

represent are connected by links into a network

• Use a frame in which related attributes are grouped together in a structure

and processed by a set of related procedures called daemons or fillers

CSCI485 Intelligent Agents 6

Neural networks (=Connectionism)

has more to do with how human or natural

intelligence occurs

based on the architecture found in the brain processes info by processing large amounts of raw

data in a parallel manner

compared to symbol processing, they perform

relatively low-level cognitive functions

knowledge gained through learning is stored in the

connection weights and is not available for examination

ability to learn from surroundings is a crucial function

CSCI485 Intelligent Agents 7

The Internet and the Web

hyper text transfer protocol brought the internet from the realm of

academia and computer technologists into the public consciousness

Knowing that the info is out there but not knowing how to find it is

• frustrating. There are popular search engines, but still web

browsing is still a hit or miss proposition.

Here intelligent agents will emerge as truly useful personal

assistants by searching, finding, and filtering info from the web.

B2B (business to business) and B2C (business to consumer)

require intelligent software to provide personalized info to perform automated negotiations, and to perform planning and scheduling functions.

CSCI485 Intelligent Agents 8

Intelligent Agents

Change of focus in the AI research to apply

basic AI to distributed systems, intranets and internet and web.

Initially the focus was limited to word

searches, info retrieval and filtering.

As more commercial transactions are

performed on networks, there is more interest in having smart agents that perform specific actions.

CSCI485 Intelligent Agents 9

Events, Conditions and Actions

How does the agent know that we want it to do something for us,

• r that it should respond to someone who is trying to contact us?

Deal with event, recognize condition and take action. Events: anything that happens to change the environment or

anything of which the agent should be aware.

Recognize Condition: after event, agent evaluates and

recognizes what it means and responds to it.

Actions: do things for us. We must trust that our agent is going

to behave rationally and in our best interest.

CSCI485 Intelligent Agents 10

Taxonomies of Agents

Agency: degree of autonomy that the

software agent has in representing the user to other agents, applications and computers.

Intelligence: ability of the agent to capture

and apply knowledge and processing to solve problems.

Mobility: ability to move between systems in a

network which it introduces additional complexity because raises concerns about security and cost.

CSCI485 Intelligent Agents 11

Processing Strategies

Reactive of Reflex Agents: respond in the event condition action

mode, do not have internal models of the world, respond only to external stimuli. They exhibit emergent behaviour, which is the result of simple agents where they share low level data rather than high level symbolic knowledge.

Deliberative or goal-directed Agents: have domain knowledge

and planning ability necessary to take actions hoping to reach a certain goal. May cooperate with other agents to achieve tasks, and may use symbolic AI reasoning techniques.

Collaborative agents: work together to solve problems. Can solve

large problems that are beyond the scope of a single agent, and allow modular approach based on specialization of agent functions or domain knowledge.

CSCI485 Intelligent Agents 12

Why use mobile agents?

reduce network load, because they move to a

system and do their work there.

• vercome network latency because resident on a

machine rather than remote.

encapsulate protocols as they move around

• perate autonomously

dynamically adapt to changes in system loading are heterogeneous are fault-tolerant since they move from a system that

is faulty

CSCI485 Intelligent Agents 13

Processing Functions

• intelligent agent attributes can be combined and applied to a specific

domain to create a new, function-specific intelligent agent.

• Interface agents: work as personal assistants, employ learning to adapt

themselves

• Patti Maes (1994): four ways that learning can occur:
• by watching over the user’s shoulder, observing what he does and imitating

him

• ffer advice or take actions on the user’s behalf and then learn by receiving

feedback or reinforcement from the user

• get explicit instructions from the user
• ask other agents for advice, and learn from their experience.
• Information Agents: (deliver useful info to user)
• help with the core problem of getting the right info at the right time.

They can be static or mobile.

• Agency and intelligence are the fundamental underlying capabilities on

which agents should be classified.

CSCI485 Intelligent Agents 14

Agents and Human-Computer Interfaces

Most natural way is through language, facial

expression and body language.

Example: by IBM new project, gaze

identification, where the user is looking on the screen.

CSCI485 Intelligent Agents 15

Using Java for Intelligent Agents

Overview of the Java Language Autonomy Intelligence Mobility

CSCI485 Intelligent Agents 16

Overview of the Java language

• bject-oriented
• Developers of java originally intended to use C++ but needed a

language that could execute on different computer chips.

• ideal for programming on the internet
• Java source code is compiled into an intermediate form called byte

codes which are stored in a class file and can be executed on any computer system than has Java Virtual Machine (JVM).

• Better than scripting language but slower than traditional languages

which are compiled directly.

• Just-In-Time compilers have developed for this purpose; they run

concurrently with JVM and determine what piece of code is called most frequently, which in turn are compiled into machine instructions on the fly so they do not need to be interpreted.

• Java applets are small java programs designed to be included in an

HTML document.

• Java is also a full function programming language that can be used to

write standalone applications

CSCI485 Intelligent Agents 17

Autonomy

Java applications are separate process and as such

can be long running and autonomous.

They communicate with other programs using

sockets.

Intelligent Agents are autonomous programs and

are always waiting for user request or change in the environment.

In java there is an event processing mechanism

used in the Abstract Windowing Toolkit (AWT) to pass user-defined events.

CSCI485 Intelligent Agents 18

Intelligence

Intelligence in intelligent agents can range

from logic to reasoning and learning.

AI has two major aspects: knowledge

representation and algorithms that manipulate them. Java can easily be used to implement them.

CSCI485 Intelligent Agents 19

Mobility

Java’s portability allows groups of java classes to be

sent over a network and then executed on the target machine.

Java Applets provide a mechanism for running Java

code remotely.

Some environments allow Java Processes to be

started, suspended and moved.

JavaBean delegation event model allows dynamic

registration of EventSources and Event Listeners.

The Java remote method invocation package allows

Java objects to call methods on other objects across a network creating distributed object applications.