Outline for today

* Who am I?
  - Jeff Kinne, assistant professor of CS
  - from Cincinnati, bachelor's from Xavier, PhD from Wisconsin
  - 3 kids, all young boys, wife is a high school math/CS teacher
  - most call me Jeff, Dr or Prof is you prefer

* What is this course?
  - AI.  What is that?  Can give some definitions - book uses (i) make
    computer act like a human, (ii) make computer think like a human, 
    (iii) make computer act rationally, (iv) make computer think rationally.
    - Can also give examples.  
      * robotics
      * game playing
      * expert systems
      * NLP
      * classification/learning
      * Watson (NLP, search, probabilistic reasoning, ...)
      * Others??
  - Elective for CS majors (grads and undergrads).  And one non-CS person
    present.
  - Note: I am not an expert in this field.  I will be learning some of this 
    with you!

* Go over syllabus...
  - take attendance, who is everyone.
  - flexibility in project assignments.  if you have one you'd like to do, 
    let me know and we might be able to count it.

* Look at projected schedule as well, assigned reading.
  - Things not on there - NLP, theorem proving, ...
  - If you want those, we could do some substitutions.  If you have a favorite
    topic let me know.

* Installing python, save hw0.py, open it, run it.
  - and play around with a few other things too.
  - and how to turn in homework.

* The Turing paper and Turing test...
  - what is it, will it ever happen.
  - note: because universe is finite, there is a computer program that 
    could pass the test, though it would be too large to fit in our 
    universe.  so at least on philosophical grounds, there should be no 
    reason to think it's not possible.
    - this is a refutation of the objection that to pass TT must think
      like a human.  the lookup table does not, but it still passes.
      Our goal would be to pass the test more efficiently though.
  - other objections...
  - other versions of the test: computer tries to convince interrogator
    that it is a woman, interrogator is not told talking to computer - so 
    easier to pass this version of TT.  Or try to convince interrogator 
    the computer is crazy, or ...
  - other version of test: Turing suggests trying to get the computer to 
    act/think like a child rather than an adult - step along the way to the 
    full TT.
  - other version of test: total TT - also need to have sensors to 
    look, listen, touch, smell, and have that as part of interaction.
  - other version of test: limit conversation to something specific - 
    current events, basic algebra, ...
  - other version: only get to ask yes/no questions (makes no difference...)
  - Watson?
  - What about having a computer be the interrogator to distinguish?  Something
    like a game then - for every computer trying to pass the TT, is there a 
    computer that can detect it?  Or, for every computer interrogator, is 
    there a computer that fools it and passes TT?
  - TT used in practice - CAPTCHA

* Other things about the paper.
  - Interesting how he describes things even back then.  Audience are people
    who may not know anything about computers.  
  - He starts by considering arbitrary machines, then decides on only allowing
    "digital computers".  Nowadays, people would assume computers right away.
  - Has to explain what a digital computer is, because his audience does not
    know.  Though even today people do not all know how computers work.

* Other thoughts...
  - Evolvability.  Human has had the benefit of a few billion years of 
    evolution.  How do we model that process and take it into account?
    How big an advantage is that to humans over computers?...

* Other thoughts?

* Reading for the weekend - see the schedule

* HW0 due when it's due, may add a problem or so.
  - For any HW, start it right away.  If you send me something early, I will
    look at it and let you know if it looks correct or not.  You can then 
    make corrections and turn it in again, as long as you turn it in before
    the due date.
    
* Use room A-008 or A-009 or A-015?

* And for next week, search.
  - Examples: 8 queens, sliding puzzle, sudoku, TSP, vacuum cleaner example
    in book, ..., lots of things can be modeled this way.
    - For each, start state, goal state(s), how move from one to 
      another, and cost of the search and maybe of the solution.
    - Doing a brute force search can be much better already than 
      trying all possible states (because there are lots of states that can 
      be ignored!).
    - Choices in terms of how to explore the state space: BFS, DFS, 
      depth-limited search, ...
    - Note: state space could be infinite.
    - Desirable properties of search: always finds a (optimal solution), 
      running time, memory space, simplicity of programming