Textbook Notes - AIMA, R&N

Chapter One

CSC447 - Spring 2010

                                            

What is AI?

                

     – Our definition

          gain better understanding of intelligence with the goal of building machines better able to serve humankind.

                

     – Other definitions

          4 quadrants

             – upper - thinking; lower - acting

             – leftmost - simulate humans; rightmost - strive for rationality

          results in 4 definitions

             – acting humanly :: Turing Test approach

                   Need: NLP, KR&R, ML, CV, robotics

                   Problems: TT based on deception; courtroom in Veterans Stadium

             – thinking humanly :: cognitive modeling approach

                   Problem: early days of AI there was confusion between computer

                             models & actual human processes

             – thinking rationally :: Laws of thought approach

                   Aristotle, George Boole, logicists

                   Problems: not easy to formalize incomplete knowledge;

                             solve in principle ≠ solve in practice

             – acting rationally :: rational agent approach

                   r.a. = one that acts so as to achieve best outcome (or best expected

                             outcome)

                   Needs: same skills as TT

                   2 advantages over others

                             more general than laws of thought approach

                             more amenable to scientific development

                   R&N claim: standard of rationality is mathematically well defined and

                             completely general

                   AIMA concentrates on general principles of r.a.'s and on components for

                   constructing them

                   Limitation: achieving perfect rationality in complicated environments is                      not feasible

                   Problems:

                             Philosophically, since humans set principles of rationality, this is a

                                      subset of thinking/acting humanly

                             Sheldon Principle - watch some episodes of Big Bang Theory

                             Justification of scientific induction - knotty, unsolved problem in

                                      philosophy of science

                             Waning influence of logical positivism, e.g., here

                

     – Advantages of our definition

          can use any of the other 4 approaches without being boxed in by philosophical

                   commitments

          have an identifiable, actionable and justifiable goal

          have a clear research agenda

          no need to prove anything; proof of the pudding is in the eating

          notice how many of the items mentioned on pp. 28-29 fall into this category

                

     – Alternative to the Turing Test

          the Touring Car Test - does it get us from point A to point B?

          we set the goal; if the goal is achieved we move on; no need to engage in

                   endless philosophical arguments over intelligence, rationality, etc.

                                            

The Foundations of AI

Philosophy

                

     – Questions

          Formal rules ® valid conclusions?

          Physical brain ® mind?

          Whence knowledge?

          Knowledge ® action?

                

     – Milestones & bumps in the road

          Aristotle - syllogisms

          Ramon Lull - can reason with mechanical artifact

          Thomas Hobbes - reasoning like numerical computation; artificial animal

          Calculators - da Vinci, Schickard, Pascal, Leibniz

          Schools of philosophy - rationalism, dualism, materialism, empiricism, logical

                   positivism

          Individuals - Descartes, Bacon, Locke, Hume, Wittgenstein, Carnap, Hempel

          Carnap - first theory of mind as computational process

          Justification of induction - unsolved problem

          Aristotle - connection between goals and knowledge of outcome

                   implemented by Newell & Simon in GPS (General Problem Solver)

                   Note: GPS ® STRIPS, bypassed by R. Brooks