ai overview reference: "artificial intelligence, a modern approach, 3 rd ed."

9
AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

Upload: elvin-atkinson

Post on 26-Dec-2015

219 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

AI Overview

Reference: "Artificial Intelligence, a Modern Approach, 3rd ed."

Page 2: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

What is AI?

Top Image: http://howtosplitanatom.com/news/119/Second Image: http://john-likes-movies.blogspot.com/2010/05/ai-artificial-intelligence.htmlThrid Image: http://engineeringyourfuture.wordpress.com/Fourth Image: http://thefuturebuzz.com/2008/11/07/social-media-is-like-the-matrix/Fifth Image: http://blog.solutionary.com/blog/bid/34219/Security-Event-Detection-Man-vs-MachineSixth Image: http://en.wikipedia.org/wiki/Data_(Star_Trek)

Page 3: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

4 Schools of thought (in AI academia)

Thinking Humanly Thinking Rationally

Acting Humanly Acting Rationally

Top Row: Measure the success of an AI program by how well it "reasons"Bottom Row: Measure the success of an AI program by how convincing it is.Left Column: Measure the success of an AI program by how human-like it is.Right Column: Measure the success of an AI program by successful (rational) it is.

Page 4: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

Acting Humanly• Turing Test– Natural Language Processing– Knowledge Rep.– Automated Reasoning– Machine Learning

• total Turing Test– computer vision– robotics

• Not much research being done here…

Alan Turing (1912-1954)

Top Image: http://www.wired.com/thisdayintech/2010/06/0623alan-turing-born/Bottom Image: http://withfriendship.com/user/kethan123/turing-test.php

Page 5: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

Thinking Humanly• Mimic the way a human thinks– To mimic we have to understand it.

• For this branch, it's not enough to produce the correct input/output as a human. • Cognitive Science:

– Linguistics– Philosophy– Anthropology– CS– Neuroscience– Psychology– etc.

• An active area of research, but not the main approach in AIImage: http://medicalphysicsweb.org/cws/article/research/33586

Page 6: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

Thinking Rationally

• Logic– e.g (a syllogism):

• Socrates is a man (a fact)• All met are mortal (a rule)• Therefore…Socrates is mortal (a new fact, deduced from the

other two)

• We could theoretically solve any problem that could be expressed in this form– With more than a few hundred facts, it's SLOWW– Not easy to express some things in this fashion.

• Prolog (programming language)

Image: http://en.wikipedia.org/wiki/Aristotle

Page 7: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

Acting Rationally

• Behaving "correctly" in the context the system is meant to work in

• The most general of all 4.– We can (but don't have to) mimic human brain processes.– We can (but don't have to) use classical logic (thinking

rationally)– We can (but don't have to) make this program act human.

• Or we can just mimic one part of acting humanly.

• This is the approach I'll use.– We're trying to implement programs with behavior that appears to be involving thought / intelligence.

Page 8: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

Which are you?

Thinking Humanly Thinking Rationally

Acting Humanly Acting Rationally

Top Row: Measure the success of an AI program by how well it "reasons"Bottom Row: Measure the success of an AI program by how convincing it is.Left Column: Measure the success of an AI program by how human-like it is.Right Column: Measure the success of an AI program by successful (rational) it is.

Page 9: AI Overview Reference: "Artificial Intelligence, a Modern Approach, 3 rd ed."

AI foundations• AI draws from a lot of different disciplines:– Philosophy (rules of logic, esp)– Mathematics (proofs, algorithm analysis, computability,

etc)– Economics (utility -- what's the best action considering

their cost and outcomes?)– Neuroscience / Biology / Psychology (many algorithms

are based on these)– Computer Engineering (we have to run the programs

somewhere…and there are limitations)– Linguistics (relationship between language and thought)– …