Transcript Intelligent Systems

Intelligent Systems
Q: Where to start?
A: At the beginning (1940)
by Denis Riordan
Reference
Modern Artificial Intelligence began in the middle of the last century.
Alan Turing proposed the question, ‘Can machines think?’
A QUARTERLY REVIEW OF PSYCHOLOGY AND PHILOSOPHY, Computing machinery
and intelligence - A. M. Turing, p.433, VOL. LIX. No.236. October, 1950
http://www.abelard.org/turpap/turpap.php
1
General References
Find some of your own!
[1] Negnevitsky M., Artificial Intelligence, A Guide to Intelligent Systems, 2011
[2] Russell S. and Norvig P., Artificial Intelligence, A Modern Approach
[3] Udacity, https://www.udacity.com/course/cs271
[4] Hodges A. , “The Alan Turing Home Page”, http://www.turing.org.uk/turing/
[5] Association for the Advancement of Artificial Intelligence, http://www.aaai.org/home.html
[6] Stone and Hirsh, “Artificial Intelligence: The Next Twenty-Five Years” ,
http://www.aaai.org/ojs/index.php/aimagazine/article/view/1852/1750
[7] IBM’s Watson Program on Jeopardy,
http://www.aaai.org/AITopics/pmwiki/pmwiki.php/AITopics/HomePage
[8] IBMWatson. http://www-03.ibm.com/innovation/us/watson/index.shtml
2
Goals of the Course
1. To understand different views of AI
2. To apply algorithms from AI to solve some real world
problems
3
Exercise
Give a (your) definition of intelligence (three lines)?
•
•
•
•
Give your
of AI (three lines)?
thinks
like definition
a human
thinks rationally
acts like human
acts rationally
4
Approaches to AI
•
•
•
•
thinking humanly – cognitive modeling
acting humanly - Turing
thinking rationally – logicians
acting rationally – achieve the best outcome
5
Class Exercise
Use your your definition.
According to your definition: Give an example of an interesting
intelligent system that you have encountered? Explain?
Can your definition be used to decide, for example,
whether an extraterrestrial
radio signal indicates an alien intelligence? See http://www.planetary.org/programs/projects/setiathome/
6
Everyday use of AI
•
•
•
•
•
•
Games
Medicine
Finance
The Web
Robotics
Give some other
7
Agent Model
•
•
•
•
•
Agent
Environment
Actuator (action)
Sensor (percept)
Performance Measure
8
Minimal Agent
9
Properties of Environments
•
•
•
•
•
•
Observability (full vs partial)
Uncertainty (deterministic vs stochastic)
Experience (episodic vs sequential)
Change (static vs dynamic)
Continuity (discrete vs continuous)
Single vs Multiagent
10
Learning Agent
11
RATIONAL AGENT
• Agent that selects an action to maximize
performance given a percept sequence and
knowledge base
12
Class Exercise
Measures of Intelligence
13
Some AI Problems in Research
• Represent medical knowledge for health
diagnostics and treatment planning
• Game Players that learn from scratch
• Encyclopedia on Demand – Produce a 5000 word
encyclopedia style article, on a given subject, by
summarizing from the relevant information on the
web
• Robot drivers – taxi
• Natural Language Understanding
14
Examples of AI Uses
in Industry
•
•
•
•
•
•
Remote Diagnostics
Healthcare, clinical guidelines and pathways
Implementing Business Rules
Data Mining
Natural language
Product selection
15
What are we trying to accomplish?
• The study of how to make agents that do
things at which, for the moment, people are
better (adapted: Rich and Knight, 1991)
16
Schools of thought?
• Success compared to human performance.
• Focus on mechanisms, structure
• Evolution
• Swarm Intelligence
17
Turing Test
• The computer should be interrogated by a
human via a teletype and passes if the
interrogator cannot tell if there is a human
or a computer at the other end
18
Turing Test: Phase 1
19
Turing Imitation Game: Phase 2

In the second phase of the game, the man is
replaced by a computer programmed to deceive the
interrogator as the man did.
20
Turing Imitation Game: Phase 2
21
The Turing test is objective
• By maintaining communication between the
human and the machine via terminals, the test
gives us an objective standard view on
intelligence.
22
Turing’s Prediction
• In 1946 Turing predicted that by 2000 a
computer could be programmed to have a
conversation with a human interrogator for
five minutes and would have a 30% chance
of deceiving the interrogator that it was
human.
23
Approaches to AI
1. Alan Turing defined intelligent behavior as the
ability to achieve human level performance in
cognitive tasks, sufficient to fool an interrogator
2. Minsky defined intelligence in terms of
mechanisms. e.g., a human is a 'meat' machine
3. More recently some scientists have come to view
intelligence as a evolutionary process - evolving
in a competitive environment. The more
competitive the more intelligent.
4. Swarm Intelligence
24
Class Exercise
How does your definition
of machine intelligence given earlier
fit in with
the four approaches given above ?
25