Greg Detre
14:30 Wednesday, September 18, 2002
Peirce 209
different goals of AI: 2x2 matrix
think like humans�������� think rationally
act like humans����������� act rationally
�Perceptrons� (Minsky & Papert): although single-layer perceptron could learn anything it could represent, it can�t represent very much
things computers can�t do
driving a cab in the centre of cairo
can�t play bridge at an expert level � a lot of uncertainty, much is unobservable, can�t see cards, large search space
go � combinatorial explosion, holistic
writing an intentionally funny story
orbitz apparently uses AI techniques
�as an analogy with a more mature science such medicine, just because the formula for the ultimate goal of �eternal youth and health� has not been achieved, one cannot argue that the field of medicine hasn�t advanced tremendously� � see link below, relating to Wired article attack on AI claims
main ideas of this course
representational
expressiveness vs computational
tractability
bounded-rationality � best possible given computational resources
exploit heuristic information and leverage structure
we will build
search-based problem-solving for the �vacuum world�
game player � connect4
constraint satisfaction solver for graph-colouring + inference
semantic network to represent + reason with knowledge
planning domains
agents
goal, environment, sensors, actuators
course outline
search
knowledge representatoin
planning
distributed AI
you don�t want to explode the number of states in your world by the time dimension � early difficult problem
robocup goal � by the year 2050 to develop a team of fully autonomous humanoid robots that can beat the world soccer champions
also a simulation league
trading agent competition � buying entertainment/flight/hotel tickets � harvard wants to enter in a couple of years� time
download lecture notes
buy �AI � A modern approach�
electronic discussion board � use it for most questions directed at TAs
class project/final paper
sections
see � www.redherring.com/insider/2002/08/ai082302.html
tacs.eecs.mich.edu
branching factor of chess � 80???