CSCI 373: Artificial Intelligence

CSCI 373: Artificial Intelligence

Andrea DanylukSeptember 6, 2013

• Who are you?– Roster

“Artificial Intelligence”

• First thing that comes to mind?

• What is “intelligence”?– According to Merriam-Webster:

• The ability to learn or understand or to deal with new or trying situations: the skilled use of reason

• The ability to apply knowledge to manipulate one’s environment or to think abstractly as measured by objective criteria (as tests)

• What [abilities] does intelligence involve?– Ability to learn– Ability to reason– Ability to apply “knowledge”– Ability to think abstractly– Ability to demonstrate the above skills by

• Taking in percepts• Acting (physically or otherwise)

Typically derive from our understanding of the most intelligent entities we know:

Ourselves.

What is AI?

Think Humanly Think Rationally

Act Humanly Act Rationally

The science of making machines that

[CS 188 UC Berkeley]

Rationality

• An ideal performance measure; does a system do the “right thing” given what it knows?

• Involves a combination of mathematics and engineering.

Goals of “Computational Rationality”

• Engineering– To solve real-world problems– To build systems that exhibit rational behavior

• Scientific– To understand what kind of computational

mechanisms are needed for modeling rational behavior

Logistics

• Where/When: Here (TPL 114) on MWF at 10am• Prof: Me (Andrea Danyluk)• Email: [email protected]• Phone: x2178• Office: TCL 305• Office Hours: Pretty much any time my office door is

open. And Mon 1:30-3:30, Tues 1:30-2:30, Thurs 1-2• 373 website:

www.cs.williams.edu/~andrea/cs373

What we’ll cover• Making Decisions

– Fast search/planning/problem solving– Adversarial search– Constraint satisfaction

• Reasoning under uncertainty– Bayes’ nets– Decision theory

• Logic• Learning

– Reinforcement learning– A bit of supervised classifier learning

Work and grading

• Programming assignments (50%)– One tutorial plus five more– Python– Teamwork (required, except for tutorial)– Autograding + code review– “The Pacman Assignments”– Do not – absolutely not – post solutions to any

part of these assignments!

Work and grading• Project (25%+10%)

– Topic of your choice (with my sign-off)– Short written proposal– Python or Java (some exceptions allowed)– Deliverables

• code+demo+presentation (25%)• Paper (10%)

• One exam (10%)– Short take-home

• Other (5%)– Short response papers– Being here, being engaged, being prepared….

Due dates and lateness

• Pacman/machine learning assignments– Up to 4 free late days (can use at most 2 at once)

• Project code+demo+presentation– 10% late penalty per day

• Paper responses and final paper– Must be submitted on time

Honor Code

• Exam: follow the College Honor Code. The work on the exam must be your own work.

• Assignments:– Listen carefully– Read the Honor Code Guidelines in the syllabus– When in doubt, ask me

What can AI do?

• Play a decent game of chess?– Play a decent game of Jeopardy?• Ambiguity of language

– After Governor Baldridge watched the lion perform, he was taken to Main Street and fed twenty-five pounds of red meat in front of the Fox Theater.

– Dr. Benjamin Porter visited the school yesterday and lectured on"Destructive Pests". A large number were present.

– Play a decent game of table tennis?– Play a decent game of soccer?• What the robot sees

[Adapted from Russell]

What can AI do?

• Drive safely at high speed?– Drive safely in an urban setting?

• Schedule and manage a fleet of luxury limousines for business travelers for one of the largest travel agencies in Hong Kong?

• Assist in making grad school admission decisions?• Identify disease outbreaks?• Monitor prescriptions?

What can AI do?

• Write a news brief?• Be a punster?– “What do you call a spicy missile? A hot shot!”– What is the difference between leaves and a car?

One you brush and rake, the other you rush and brake.