First
Year Seminar
The Limits
of Machine Intelligence
Syllabus
(Fall, 2003)
Professor:
Dr. Joseph T. Wunderlich
Office: Nicarry 244
Phone: 361-1295
Email: wunderjt@etown.edu
Office Hours: http://users.etown.edu/w/wunderjt/schedules/f03schedule.html
Objectives:
Machine Intelligence is found in many modern-day technologies and can be
defined as encompassing all of the developments in both
symbolic artificial intelligence and artificial neural networks. Traditional
symbolic AI uses programmed heuristics and forms of knowledge representation to
produce results in a seemingly more intelligent way than typical computer programs.
Artificial neural networks are a form of connectionist computer architecture
where many simple computational nodes are connected in an
architecture similar to that of biological brains for the purpose of
solving problems which require rapid adaptation or solutions where underlying
governing equations are not known or cannot be easily computed. This course
explores the limits of machine intelligence by comparing the potential of these
man-made systems to the known “mental ability” of common biological life forms.
The discussion begins with a study of basic animal abilities such as
adaptation, self-preservation, motor-coordination, and processing complex
sensory information. More advanced abilities are then explored including
abstraction, tool-manipulation, creativity, and emotional expression. Ethical
issues are also discussed.
Course
Credit: Three (plus one for required concurrent
colloquium)
Course Text:
Supplimental
Grading:
1.
Do
something for “into the streets” on
2.
Help
as an usher on a Wednesday at
3.
Propose
something (must have my approval before you do it)
COURSE GRADE:
(60-62)=D-,
(63-67)=D, (68-69)=D+, (70-72)=C-, (73-77)=C, (78-79)=C+, (80-82)=B-,
(83-87)=B, (88-89)=B+, (90-92)=A-, (93-100)=A
(with
any fractional part rounded to the nearest integer)
Attendance:
Exams will primarily cover material presented in lecture -- some of which may
not be found in the course text or supplemental readings.
Academic Honesty:
Elizabethtown College Pledge of Integrity: "Elizabethtown College is a
community engaged in a living and learning experience, the foundation of which
is mutual trust and respect. Therefore, we will strive to behave toward one
another with respect for the rights of others, and we promise to represent as
our work only that which is indeed our own, refraining from all forms of lying,
plagiarizing, and cheating."
Semester Research Paper:
Grading: PROPOSAL:
5% of total course grade; letter graded
Due Date: To
be announced
Late Penalties: 5%
per class period
Grading: FINAL
PROJECT: 25% of total course grade; letter graded
Due Date: To
be announced
Late Penalties: 20%
per class period, up until last day of class
The
oral proposal should take approximately five to seven minutes (not including
answering audience questions), and must contain visuals (using any software or
medium you wish). Some things to avoid in your presentations:
v More
than 30 words per visual.
v Reading directly from a
script.
v Poor contrast between
text and background.
v Too many sound effects
(e.g., screeching car for every bullet).
v Too many slides for
allotted time (e.g., more than 3 slides per minute).
v Speaking monotonically.
v Never making eye contact
with audience.
·
A
good presentation:
o
Is as visual as possible. If a picture is worth a thousand words, an
equation or graph is worth 10,000.
o
Often has an image on every page (e.g., clip-art,
photo, animation, etc.) which is an abstraction of the subject matter on the
slide (i.e., invokes an idea).
o
Has
a clear objective (e.g., to entertain, to sell, to motivate, or to report
findings).
o
Has
a good “opener” (e.g., an agenda, a
quotation, a question, or a declaration)
o
Is
organized clearly and logically (e.g., by problem then solution; or by
priorities – least-to-most or most-to-least).
o
Has
the audience’s expectations understood (e.g., provide meaning and/or
motivation).
o
Minimizes unnecessary details (i.e., don’t overwhelm
audience with too much info).
o
Has
good transitions between main points
(i.e., short, attention-getting)
o
Has
a good “closing” (i.e., summarizes
main ideas, restates purpose of presentation)
o
Is
flexible (i.e., can be modified on the fly if questions are allowed during
presentation)
All projects must relate to machine
intelligence, and the project should be mostly a research project. However if
you are sure you have the expertise, you may build something.
On
the project due-date, both written and oral reports are due. The oral report
must be done using either PowerPoint or a web page created by you for the project. It should take approximately ten
to fifteen minutes and contain an appropriate number of visuals. The written
report must be in two-column,
single-spaced, 10-point font and must use the formatting dictated by the paper:
“Defining the limits of machine
intelligence” by
Dr. Wunderlich (to be handed out in class, and electronically). Also,
attach to your paper the accompanying PowerPoint presentation printed six
slides per page (or screen shots of your web page presentation). The paper should be between 4 to 6 pages
and include:
1) An abstract (one or two paragraphs)
2) Miscellaneous discussions of details (this
depends on the type of project)
3) Conclusions
4) A bibliography (i.e., a list of referenced
material) – call it “References”
5) An appendix containing schematics, manufacturers literature, etc.
Course Outline:
I.
Neural
Networks vs. Symbolic Artificial Intelligence
A) “Bottom-up” Brain Models
B) “Top-Down” Brain Models
C) Evolution of Neural
Networks
a)
Perceptrons
b) Learning Rules
c)
Underlying
Neural Network Mathematical Theories
d) Neural Network
Applications
e)
Neural
Network Hardware and Software
D) Symbolic Artificial
Intelligence
a)
Predicate
Calculus
b) Knowledge Representation
c)
A.I.
Programming Languages
1.
Prolog
2.
LISP
d) Expert Systems
II.
Biological
vs. Machine Intelligence. The following “Mental Abilities” will be discussed by
answering five simple questions:
(1) What can humans do?
(2) What can a simple insect do? (e.g., a
spider)
(3) What can a conventional computer do?
(4) What can symbolic A.I. programming do?
(5) What can Artificial Neural networks do?
A) Basic Animal Abilities:
Acquire and retain knowledge, Solve problems, Motor coordination, Acquire
energy, Protect self, Sensory processing, Real-time thought React
instinctively, Anticipate, Predict, Communicate, Generalize, Associate,
Recognition patterns, Robust under partial failure, Autonomous thought, Drive
to reproduce, Stability, Repeatability, Predictability, Multitask
B) Complex Abilities:
Abstraction, Intuition, Common sense, Manipulate tools, Heuristics, Inference,
Hypothesis testing, Self-discipline, impulse-control, Ethical behavior,
Selective awareness, Open to inspection, Emotions, Imagination, Creativity,
Passion, Playfulness, Empathy, Leadership, Self-awareness, Awareness of mortality,
Group psychology
III.
Merging
Neural Networks with Symbolic Artificial Intelligence
IV.
Artificial
Humanoids
A) Emulating Human
Physiology
a)
Biomechanics
b) Senses
c)
Control
Systems
V.
Ethical
Issues involving Machine Intelligence
A) Replacing Humans
B) Aiding Humans
C) Military Uses