Artificial Intelligence Toolbox Part 1: How to find solutions Myra Wilson e-mail

Artificial Intelligence Toolbox
Part 1: How to find solutions
Myra Wilson
e-mail [email protected]
Aberystwyth University
Commitment
●
18 lectures
●
6 hours practical
●
●
Rest of time spent on background reading
and assignment (76 hours!)
Assessment:
–
Exam 70%
–
Assignment 30%
Course Content
●
Introduction - 2 lectures – mxw
●
Search – 6 lectures – mxw
●
●
●
Evolutionary Computation for Search and
Optimisation – 3 lectures – rkj
Clustering – 4 lectures – rjk
Case-based reasoning and k-nearest
neighbour – 3 lectures – rkj
What you need to do....
●
●
●
●
Attend lectures – slides are not enough.
They are only pointers to the subject
Do the practicals – there may be exam
questions which these help with
Read the books – buy, borrow... many AI
books in the library. Read around the
subject
Access the web – lots of stuff out there
Book List
●
●
●
●
Russell,S. and Norvig,P. - Artificial Intelligence : a
modern approach, 3rd edn, Prentice Hall, 2010.
ISBN 0-13-207148-7
Cawsey,A. - The essence of artificial intelligence,
Prentice Hall, 1998.
Ginsberg,M. - Essentials of artificial intelligence,
Morgan Kaufmann, 1993.
Coppin,B. - Artificial Intelligence Illuminated,
Jones and Bartlett Publishers, 2004. ISBN 0-76373230-3
What is Artificial Intelligence?
●
Understand intelligent entities
●
Build intelligent entities
●
Study constructed intelligent entities
Intelligent Systems
How is it possible for animal/electrical
brains to
–
Perceive?
–
Understand?
–
Predict?
–
Manipulate?
It can be done.........
We have the proof!
●
Computers and robots
can be used as tools to
test theories
What is Artificial Intelligence?
●
●
AI is often burdened with over-promising
and grandiosity
The gap between AI engineering and AI as
a model of intelligence is so large that
trying to bridge it almost inevitably leads to
assertions that later prove embarrasing.
McCarthy said AI was “the science and
engineering of making intelligent machines”
AI – Hamid Ekbia
●
Artificial Intelligence seeks to do three
things at the same time:
–
as an engineering practice, AI seeks to
build precise working systems;
–
as a scientific practice, it seeks to explain
the human mind and human behaviour;
–
as a discursive practice, it seeks to use
psyychological terms (derived from its
scientific practive) to describe what its
artifacts (built through the engineering
practice) do.
What is Artificial Intelligence?
1. ``[The automation of] activities that we associate with
human thinking, activities such as decision-making,
problem solving, learning...'' (Bellman, 1978).
2. ``The study of the computations that make it possible to
perceive, reason, and act'' (Winston, 1992).
3. ``The study of how to make computers do things at which,
at the moment, people are better'' (Rich and Knight,
1991).
4. ``The branch of computer science that is concerned with
the automation of intelligent behaviour'' (Luger and
Stubblefield, 1993).
Definitions of AI
Better
understanding
of
human/animal
intelligence
Emulate and
model
human/animal
behaviour
1.
Think like
humans
3.
Act like
humans
2.
Think
rationally
4.
Act
rationally
Develop
inspired
techniques
which lead to
smarter
programs and
machines.
Optimise
performance.
Foundations
●
Philosophy
●
Mathematics
●
Psychology
●
Computer Engineering
●
Linguistics
●
Biology
Alan Turing – Turing Test
●
See http://www.turing.org.uk/turing/
●
See also http://en.wikipedia.org/wiki/Turing_test
●
Proposed by Alan Turing (1950) to provide a
satisfactory operational definition of intelligence
The Turing Test
Turing Test
A computer passes the test if a human
interrogator, after posing written
questions, cannot tell if the response is
from a human or a computer
To pass the Turing Test?
●
natural language processing
●
knowledge representation
●
automated reasoning
●
machine learning
See chatterboxes such as Eliza, Alice etc
–
http://www.dmoz.org/Computers/Artificial_Intelligence/
Natural_Language/Chatterbots/
The Turing Test
●
●
Assumes that intelligence is defined by
human-like reasoning and communication
Is this the case?
The Chinese Room
●
●
●
Searle argued that behaving intelligently was not
enough (1980's)
He devised a thought experiment called the
“Chinese Room”
http://www.iep.utm.edu/chineser/
What was the Chinese Room?
●
●
●
●
You have a rule book that allows you to look up
Chinese sentences although you do not speak
Chinese
The book tells you how to reply to them in Chinese
You can then behave in an apparently intelligent way
copying replies onto stacks of paper
He claimed that although they appeared intelligent,
computers would be using the equivalent of a rule
book
–
the rule book and stacks of paper, just being
paper, do not understand Chinese
What did this say?
●
●
From the outside, it looks like the computer
is “intelligent”
Running programs does not necessarily
generate understanding
–
“Weak AI” claims computers just simulate
thought
their understanding is not real
● machines just act as if they could think
“Strong AI” claims that machines are actually
thinking
●
–
●
●
Classics
Games and AI
–
Noughts and Crosses
–
Chess – Deep Blue 1997
–
Connect 4, Othello, Backgammon,
Scrabble, Bridge.....
Current
–
RPG/Adventure
–
Strategy/Tactical/Combat
–
Racing
–
Artificial Life (Creatures, Spore)
What AI do they use?
●
Pathfinding/Search
●
Genetics
●
A-life
●
Neural Nets
●
Fuzzy Logics
●
State Machines (condition, event, action)
●
Agents.....
AI and Games
“In the next 20 years, AI is going to be what
differentiates one game from another”
–
Finite State machines used to control
character behaviour
–
A* used to plan paths (very common)
Game Techniques
Given the state of the world, what do I do
next?
●
Already in games:
–
●
Finite State machines, decision trees, rule
based systems, neural networks, fuzzy
logic
In the academic world (too slow for real
world)
–
Complex planning, logic programming,
genetic algorithms, Bayes nets
Challenges
●
Must be smart but purposely flawed
●
No unintended weaknesses
●
Must perform in real time (CPU)
●
Configurable by designers
●
Amount and type of AI for games can vary
AI Case Study - Creatures
What was “Creatures”?
●
Computer game published in 1996 by Mindscape
●
Creator Steve Grand
●
http://www.gamewaredevelopment.co.uk/creatures_index.php
●
http://en.wikipedia.org/wiki/Creatures_(artificial_life_program)
●
Based on a creature called a “Norn”
●
At one point, about 5 million norns worldwide!
●
Intended as “virtual pets”
●
Inspired by animal biology
The norn world
Norns
●
Composed of neurons, genes, biochemicals etc
●
The “gene” dictates how an organism is made
●
●
●
●
The behaviour emerges from the interaction of the
parts rather than explicit programming
Norns can learn about their environment
Norns can form relationships and produce offspring
which inherit structures from their parents
They can become ill
How did they do this?
●
●
●
Each norn has a neural network responsible for
sensory-motor coordination and motor control
“artificial biochemistry” which models a simple
energy metabolism
“hormonal” system which interacts with the neural
net to model neuronal activity and development
●
Hebbian learning mechanism allows adaptation
●
Genetic encoding allows evolution
Creatures
●
Interactive entertainment environment
●
Created synthetic biological agents
●
Sufficiently life-like to sell well :-)
●
Observer's tendency to anthropomorphism!
●
Approx 320 interacting genes
●
Engaged public with concept of Artifical Life
AI Problems
●
●
●
Formal tasks – playing chess or card
games, solving puzzles, mathematical or
logic problems.
Expert tasks – medical diagnosis,
engineering, scheduling, computer
hardware design.
Mundane tasks – everyday speech, written
language, perception, walking, handling.
How do we solve them?
That's what this course is all about!
–
Search
–
Evolutionary Computation
–
Clustering
–
Case based reasoning and k-nearest
neighbour
Summary
●
AI takes its inspiration from many different
disciplines
●
Acting and thinking, human and machine
●
Turing test
●
Games