Download 1. W01-Definition

Lecture Notes
Artificial Intelligence:
Definition
Dae-Won Kim
School of Computer Science & Engineering
Chung-Ang University
What are AI Systems?
Deep Blue defeated
the world chess
champion Garry
Kasparov in 1997
During the 1991 Gulf War, US forces
deployed an AI logistics planning
and scheduling program that
involved up to 50,000 vehicles,
cargo, and people
Proverb solves crossword puzzles
better than most humans
Sony’s AIBO and Honda’s ASIMO
Web Agents & Search engines:
Google, Yahoo
Recognition Systems: Speech,
Character, Face, Iris, Fingerprint
Virtual Reality and Computer Vision
Potted History of AI
1943
1950
1950s
1956
1965
1966
1969
1980
1988
1985
1988
1995
2000
2003
McCulloch & Pitts: Boolean circuit model of brain
Turing’s “Computing Machinery and Intelligence”
Early AI programs
Dartmouth meeting: “Artificial Intelligence” adopted
Robinson’s complete algorithm for logical reasoning
AI discovers computational complexity
Neural network research almost disappears
Early development of knowledge-based systems
Expert systems industry booms
Expert systems industry busts: “AI Winter”
Neural networks return to popularity
Resurgence of probability, soft computing.
Agents, agents, everywhere … with Data Mining
Bioinformatics powered by Human Genome Project
Human-level AI back on the agenda: challengeable
Some researchers consider AI as
one of the four concepts:
1. Systems that think like humans
2. Systems that think rationally
3. Systems that act like humans
4. Systems that act rationally
AI: Acting humanly
Turing (1950): “The Turing Test”
Can machines think?
Can machines behave intelligently?
Turing test is The ‘Imitation’ Game
Predicted that by 2000, a machine
might have 30% chance of fooling a
lay person for 5 min.
In 2014, something has happened.
http://www.bbc.com/news/technology-27762088
Problem: Turing test is NOT …
Turing test is NOT reproducible and
amendable to mathematical analysis
AI: Thinking humanly
It requires scientific theories of
internal activities of the brain
What level of abstraction?
“Knowledge” or “circuits”.
How to validate? Requires something
Requires: Cognitive Science
Predicting and testing behavior of
human subjects (top-down)
Requires: Cognitive Neuroscience
Direct identification from
neurological data (bottom up)
Problem: Thinking humanly is NOT
Both are distinct from AI in CS
The available theories do not
explain anything resembling
human-level general intelligence.
AI: Thinking rationally
Laws of Thought: “What are correct
arguments/thought processes?”
by Aristotle
Several Greek schools developed
various forms of logic:
Logic: notation and rules of
derivation of thoughts
Problem: Thinking rationally is NOT
Not all intelligent behavior is
mediated by logical deliberation
AI: Acting rationally
Rational behavior:
doing the RIGHT thing
The RIGHT thing:
that which is expected to
maximize goal achievement,
given the available information
An agent is an entity that
perceives and acts.
Agents include humans, robots,
programs, systems, etc.
This course is about designing
rational
agents/SWs/programs/platforms.
Abstractly, an agent is a function
from percept histories to actions
f:PA
The agent program runs on the
physical architecture to produce f
For any given class of tasks and
environments, we seek the agent
with the best performance.
Problem: Acting rationally is NOT
Computational limitations make
perfect rationality unachievable
e.g.) NP-hard problems
Design best program for given
machine resources
Which of the following can be done at present?
•
•
•
•
•
•
•
•
•
•
•
Play a decent game of table tennis
Drive safely along a curving mountain road
Drive safely along Telegraph Avenue
Buy a week’s worth of groceries on the web
Discover and prove a new mathematical theorem
Design and execute a research program in biology
Write an intentionally funny story
Give legal advice in a specialized area of law
Translate spoken English into Swedish in real time
Perform a complex surgical operation
Converse successfully with another person for an
hour
Artificial Intelligence
Intelligent Agents
Dae-Won Kim
School of Computer Science & Engineering
Chung-Ang University
The agent function maps from
percept histories to actions:
f:PA
A Vacuum-cleaner Agent
• Perception: ?
• Actions: ?
• Perception: location and contents [A, Dirty].
• Actions: Left, Right, Suck, NoOp
Problem: A Vacuum-cleaner Agent
What is the right function?
Let’s talk about Rationality
A rational agent chooses whichever
action maximizes the expected
value of the performance measure
given the percept sequence to date
What is performance measure?
1 point per square cleaned up in
time T?
Minus 1 point per move?
Penalize for > k dirty squares?
Therefore, we can say
Rational  omniscient
Perception may not supply all
information
Rational  clairvoyant
Action outcomes may not be as
expected
Hence, rational  perfect
To design a rational agent, we must
specify the task environment (PEAS)
• Performance measure
• Environment
• Actuators
• Sensors
Consider the task of designing the
Google driverless car
• P: safety, comfort, profits, legality
• E: streets, freeways, traffic, weather
• A: streering, accelerator, break
• S: velocity, GPS, engine sensors
Consider the task of designing an
automated internet shopping agent:
e.g., Recommender system
• P: price, quality, efficiency
• E: WWW sites, vendors
• A: display to user, follow URL
• S: HTML, XML pages
Agent Types: four basic types in
order of increasing generality
• Simple reflex agents
• Reflex agents with state
• Goal-based agents
• Utility-based agents
Simple Reflex Agents
1. If a student sleeping, then assign a penalty.
2. When applied to Vehicle driving?
Reflex Agents with State
1. Check the student’s academic history, habits..
2. Vehicle driving ?
Goal-based Agents
1. Consider goals: be a good professor in AI class
2. Vehicle driving ?
Utility-based Agents
1. Utility: performance measure
2. How good grade I will assign / Prof. I could be.
Learning Agents
All agents will be turning into learning agents.