Vous êtes sur la page 1sur 8

TG4.

1 Introduction to Intelligent System


1. Intelligent System: Describes the various commercial app. of artificial intelligence.
2. Artificial Intelligence
Subfield of computer science that studies the thought processes of humans and
re-creates the effects of those processes via machines (eg: computers and robot)
- Ultimate Goal: Build machines that can mimic human intelligence
- A widely used test to determine whether a comp. exhibits intelligent behavior
was designed by Alan Turing Turing Test: Propose a scenario in which a man and
a comp both pretend as women (men), and a human interviewer has to identify
which is the real human.
3. The major categories of intelligent system:
- Expert system
- Neutral network
- Fuzzy Logic
- Genetic Algorithms
- Intelligent Agents
4. NI VS. AI
Capabilities

Natural Intelligence

Artificial Intelligence

Preservation of knowledge

Perishable from an org. point


of view
Difficult,
Expensive,
Takes Time

Permanent

Duplication of and
dissemination of knowledge

Total cost of knowledge

Document ability of Process


and Knowledge
Creativity
Use of sensory exp.

Can be erratic and in


consistent,
Incomplete at times
Difficult,
Expensive
High
Direct and rich in possibilities

Recognizing patterns and


relationship

Fast,
Easy to Explain

Reasoning

Making use of wide context


of experience

Easy,
Fast, and
Inexpensive once in a
computer
Consistent and
Thorough
Fairly Easy,
Inexpensive
Low, Uninspired
Must be interpreted first;
Limited
Machine learning still not as
good as people in most case,
But in some cases better
than people
Good only in narrow,
Focused,
Stable Domains

5. Intelligent System Example:


- UJam.com Double as a studio band & Recoding Studio
Analyze melody and produce sophisticated harmonies, bass lines,
drum track an horn parts.
- X-Ray A Murky Puzzle

CT/ MRI
NEW HR Image (X-rays, MRI scan, Ultrasound, CT Scan)
Group together biological structure that share hard- to detect
similarities
Measure Tissue Density (Color Code Tissue Similarity)
- Google Goggles
Snap and Analyze the Product
- Princeton Locomotive and Shop Management System (PLASMA)
Uses algorithms to analyze the railroad operation by tracking thousands
of variables to predict the impact of change in fleet size, maintenance
policies, transit time and other factors.
Used vast amount of historical data.

TG4.2 Expert System


1. Expert Systems (ESs) - are computer system that attempts to mimic human experts
by applying expertise in a specific domain.
- Support Decision-making OR Replace them
* IBMs Watson: 2007, IBM scientists have been developing what
was expected to be worlds most advanced
question- answering sys (Watson)
Watson can understand everyday human
language. & supply a precise, factual, correct answer.
Watson was input with a lot of knowledge
database (X online)
Watson uses more than 100 algorithms to analyze
a question in different ways and generate hundred of
possible answer.
***Watson thinks in probability***
Mid 2011, IBM was training Watson in Medicine
with medical textbooks and medical question etc.
Multiple choice might help a doctor from anchoring
/ getting too attached to particular diagnosis
(Jeopardy TV Shows)
IBM envisioned several uses for medical Watson :
Allow doctor to connect to Watson by
speaking into hand-held device (Speech
recog and Cloud)
Serving as a repository
Provide as second opinion
* HR Department Assigning score for each candidate (lessen
work load for HR Manager)
* Microsoft Troubleshooting software
2. Transfer expertise from Expert to Comp, involved four activities:

- Knowledge acquisition
- Knowledge representation
- Knowledge Inference
- Knowledge Transfer
3. The components of Expert Sys. (ES)
- Knowledge Base: Knowledge necessary for understanding,
formulating & solving problem.
Fact: Problem Situation
Rules: Direct the use of knowledge to solve
specific problem
- Inference Engineering: A computer program that provides a
methodology for reasoning and formulating
conclusion. (Brain of ES)
- User Interface: Enable user to communicate w/ computers
Natural Language, Question-Answer Format,
Graphics
- Blackboard (workplace): An area of working memory set aside for
the description of a current problem. (Input /
Database)
- Explanation Subsystem (Justifier): Recommendation
Knowledge Refining Sys to analyze own
performance.
3. Ten Generic Categories of ES:
- Interpretation (Inferring from obs)
- Prediction (Inferring consequences)
- Diagnosis (Inferring Malfunctions from obs)
- Design (Configuring obj under constraint)
- Planning (Dev plans to achieve goal(s))
-Monitoring (Comparing obs to plans, flagging exception)
- Debugging (Prescribing remedies for malfunction)
- Repair (Executing a plan to admin a prescriber remedy)
- Instruction (Diagnosis, Debug, and Correcting Student performance)
- Ctrl (Interpreting, predicting, repairing and monitoring sys behavior)
4. Benefits of ES:
- Increase output and productivity
- Increase quality
- Capture dissemination of scarce expertise
- Operations in hazardous environment
- Accessibility to knowledge and help desks
- Reliability
- Ability to work with incomplete and uncertain info.
- Provision of training
- Enhancement of decision-making and problem- solving capabilities
- Decrease decision-making time

- Reduce Downtime.
5. Difficulties of ES:
- Transferring domain expertise from Human Expert to ES can be
difficult.
PPL cannot always explain how they know and what they know.
- Automating Process is difficult to imply
Problem is too complex; Too many Rules; Too Vague
- Potential Liability
It can arise if a business decision driven by an ES harms someone
financially.

TG 4.3Neutral Network (NN)


1. Neutral network: A sys of prog and data structure that simulates the underlying
concepts of the biological brains.
Required a large # of processor
NN is initially Trained or Fed Large amount of data and rules about
data relationship.
Included: Airline security, infectious control, combating fraud (Credit
Card), Health Care, and Telecommunication Industry.
* The Nuclear Facility in Ontario X-ray check pt. can detect weapons
and identify biological dangerous liquid.

* Diseases Alzheimers, Parkinsons, epilepsy (Build robot to


stimulate Rats)
The researcher can study the brain function and its reaction
to stimuli.
* Investors Stock Index Features, Currencies, Natural Gas and Oil
Stocks, T- Bonds, Gold Stock, other major investment.
* Bank Fraud of credit card transaction, insurance claims, fight crime,
and gauge customer satisfaction.
2. Network has three levels of interconnected nodes:
- Input (Age, Income, Occupation)
- Middle (Data key)
- Hidden (Adjust both mortgage amount and interest rates)

TG 4.4 Fuzzy Logic (FL)


1. Fuzzy Logic: a branch of mathematics that deals w/ uncertainties by simulating the
processes of human reasoning.
FL is not only able to determine the Black or White, but also the gray
area.
* Acct FL analyze assets expressed in imprecise values (Goodwill)

* Google Determine relevance of the topic that we phrased.

TG 4.5 Genetic Algorithms (GA)


1. Genetic Algorithms: Mimics the evolutionary, survival of the fittest process to
generate increasingly better solution for problem. (Produce the
best output)
- Three functional characteristics:
Selection (survival of the fittest): To give preferences to better
and better outcomes.
Cross over: Combining good outcome to create a better outcome.
Mutation: Randomly trying combination and evaluating (success
or failure) of an outcome.
* Boeing, Honda and Rolls Royce Use GA to design their vehicle.
* Mark and Spencer To manage their inventory
* Liquide (Industrial Gas) Production schedule and distribution
point in SC.

TG 4.6 Intelligent Agent (IA) (aka. BOTS)


1. Intelligent Agent: A software program that assists you, or acts on your behalf, in
performing repetitive computer- related tasks. (Use ES and FL)
* Clippy in Microsoft Word to give help.
* BotSpot and SmartSpot
2. Three types of agents:
- Information Agent: Search for info and display it to users.
The best well-known Information Agent is Buyer Agent
(AKA Shopping Bots)
Help them to find the G&S that user need on Website.
* Amazon: Display items that you might like based on
past purchased.
* Google and Ask: Google Bots Google Index Google
Repository Google Page Rank algorithms. (Search
engine process)
* The Federal Electronic Research and Review Extraction
Tool (FERRET): Jointly by Census Bureau and the Bureau
of Labor Statistic to find info. (Include employment,
health care, race )
- Monitoring and Surveillance Agent: AKA Predictive Agent
Keep observe and report some items of
interest.
*Allstate Insurance: Every 5 sec. detects
1200 data pts. It can predict 45 mins before
it crashed.

* Watch your competitor price chg. and


special offer.
* Monitor internet, discussion groups, mailing
list for stock manipulation, insider trading
and rumors that affect stock price.
A price change for on desired products
(Air ticket)
- User Agent: AKA Personal Agent
Take action on your behalf
* Check email (sort according priority rules and alerts)
* Automatically fill out form on the web.

Vous aimerez peut-être aussi