Transcript November 1 ppt. - University of Alberta

Artificial Intelligence
(and Games)
Jonathan Schaeffer
Department of Computing Science
University of Alberta
[email protected]
GAMES Research Group
• Largest AI research
group using games
• Classic Games
• Chess, checkers, go…
• Poker, hearts, spades…
• General game playing
• Commercial games
• Role-playing (BioWare)
• Sports (Electronic Arts)
What is AI?
•
•
•
•
•
Amnesty International?
Artificial Insemination?
Army Intelligence?
Air India?
Three-toed sloth from
South America?
Artificial Intelligence
• Field of computing science concerned with
making computers do tasks that we regard
as requiring intelligence.
• All encompassing: reasoning, vision,
language, learning, creativity, robotics,
planning, etc.
The Computer (R)Evolution
Need to re-think what it means to think.
AI and History
One of the most profound contributions of
the 20th century is the realization that
intelligent behavior can be realized by
non-human information processing
architectures.
AI is a Moving Target
• Imagine going back in time…
• 1910: “Black box” that can correct your
spelling and grammar
• 1960: Flying an airplane without human
intervention
• No one thinks of the above as AI
• Once something becomes understood, it
no longer is regarded as AI
Different Architectures…
…Different Solving Methods
How?
Why?
What?
Where?
When?
Looking Under the Hood
Peek inside a “smart” program
and you may find little that is
“intelligent”.
Yet computers do “intelligent”
things.
Is this a paradox?
What is intelligence?
Chess: Man vs Machine
Garry Kasparov versus DEEP BLUE (1997)
Man Versus Machine (1997)
Kasparov
Deep Blue
5’10”
6’ 5”
Height
176 lbs
2,400 lbs
Weight
34 years
0.5 years
Age
50 billion neurons Computers
512 processors
2 pos/sec
Speed 200,000,000 pos/sec
Extensive
Primitive
Knowledge
Electrical/chemical Power Source
Electrical
Enormous
None
Ego
Crossword Puzzles
• PROVERB uses multiple agents
• 34% of clues/answers are repeats
• specialized solvers (puns, word plays)
• specialized Internet searchers
• PROVERB scores well on the NY Times
puzzles without understanding the clues
• Information retrieval without understanding
the information is a powerful technique!
My Definition?
Artificial intelligence
creates the illusion of intelligent behavior
… Artificial Stupidity
Of course, some
humans try to
create the illusion
of intelligence too!
1. Most Visible AI?
• Computer games
•
•
•
•
Creating “realistic” behavior
Creating “realistic” conversations
Real-time decision making
Planning
• Façade, the first game to
evoke strong emotional
responses
Still a Long Way to Go
2. Most Successful AI?
• Building strong game-playing programs
• University of Alberta landmarks
•
•
•
•
•
•
Checkers (1994; 2007)
Chess (1997)
Othello (1997)
Scrabble (late 1990s)
Backgammon (early 2000s)
Poker, 2-player limit (2007)
3. Most Used AI?
• Email spam filters (learn patterns)
4. Biggest AI Systems?
• Visa
• Detecting potential credit card fraud
• Walmart
• Predicting buying patterns
• US Military
• Numerous applications including autonomous
vehicle navigation, data analysis, missile
guidance systems, and face recognition
5. Biggest AI Investors?
• Search engines (Google, Yahoo, Microsoft)
• Includes language translation
• Help systems (Microsoft)
• User preferences (Amazon.com, Netflix)
“A breakthrough in machine learning would
be worth ten Microsofts” Bill Gates
6. Cool AI Applications?
• Navigation
• Search and Rescue
• Mars Rover
• Deep Space 1
Automated Driving
Serious Games
• Medical: training doctors
• Military: training commanders
• Therapy: relieving combat stress
Home Robotics
• Vacuuming
• Floor washing
• Pool cleaning
• Gutter cleaning
• Mowing the lawn • Shop sweeping
Home Robotics: Tomorrow?
Sadly, not in my lifetime,
but likely yours
7. Ethical Issues?
• Love and sex with robots?
• Do robots have rights?
• Computers as judge and jury?
Coming Soon
• Many of the following exist today, but will
become pervasive
•
•
•
•
•
•
Personalized health care
Automated vehicles
High-quality language translation
Non-industrial robots
Conversational interfaces
Turning data in knowledge
Impediments
• Public perceptions from the Sci-Fi literature
• AI is largely invisible
• Cross-cutting technology with wide applicability
• “AI” inside
• Lack of knowledge in industry
• UofA produces many AI graduates each year
• Most leave the province/country to find a job
Conclusions
• AI is a major part of your life…
and you don’t even know it
• AI will become a more important part of
your life…
and you won’t even know it
• Many AI technologies are mature are
ready for you to use
Thanks to Peter Stone, Maria Cutumisu, and various Internet sites.
Thank You!
Questions?