CSC 191/291 Course Description
Can machines show self-awareness? Could robots and other machines have
subjective sensory and emotional experiences? Could they have a moral
sense? These questions are becoming increasingly relevant in our age of
increasingly intelligent machines. We will explore these issues through
readings from the AI literature as well as from philosophy and cognitive
science. Homework will include written answers to questions, essays, Lisp
exercises and -- for 291 students only -- a team programming project intended
to demonstrate the possibility of some degree of self-awareness in a computer.
The course is optionally available for writing credit. (4 hours, Spring).
191 Prerequisites: a programming course; sophomore or higher standing
(or consent of instructor); a prior course in AI, cognitive
science, linguistics, or philosophy is an advantage.
291 Prerequisites: CSC 173 and one of 200, 240, 212(HCI), 242, 280, or 282.
(See comments below.)
Lectures:
Tuesday and Thursday, 11:05p.m. to 12:20 p.m., Online Room 20
Comments on Prerequisites
Students should meet the prerequisite requirements (sophomore standing
and some programming experience--minimally the equivalent of CSC 108 or,
better, CSC 161 or CSC 171--for 191. For CSC 291 students the requirement
of having taken at least one of CSC 200, 240, 212, 242, 280, or 282 will ensure
that they can handle the "self-aware system" team programming assignment.
Students are advised not to undertake this course lightly. It is expected
to be a significant amount of work. The course has an extensive reading
list, and requires getting a broad grasp of philosophical, cognitive,
and AI issues pertaining to machine consciousness and self-awareness,
and includes a programming component.
For CSC 291 students, homeworks and the programming project will be evaluated
on the assumption that these students have previously attained some
programming maturity, and have previously been exposed to basic AI
topics such as computer vision, knowledge representation, problem
solving, language understanding, etc. However, students who have taken
the undergraduate problem seminar CSC 200 or a theory course such as CSC 280
or CSC 282 probably have sufficient maturity in the more formal aspects
of computer science to be able to make up for lack of a prior AI course.
The projects may well require the use of multiple source files
and libraries, so experience with basic linux commands, and web
utilities such as overleaf for document developmet and
github for code development is an asset.
Goals of the course
- Know how to represent basic self-knowledge in a machine (using Lisp),
and enable QA, as required for overt self-awareness
- (Especially 291 students) Develop skills in self-aware agent design,
symbolic programming
- Understand (as far as current knowledge allows) the meaning of
"consciousness" (both phenomenal & access consciousness,
including self-awareness)
- Develop an informed, reasoned view of the basic issues concerning
consciousness, in humans, animals and robots (from philosophical,
cognitive, and AI perspectives)
- Acquire basic understanding of the biological basis of consciousness
and emotion in people (& animals)
- Acquire basic understanding of AI/robotics efforts to endow machines
with consciousness and emotions
- Hone skills in philosophical and scientific writing/argumentation
- Stir up interest in one of the most intriguing topics of all time
-- which is now being explored in new ways (neuroscience & AI)
Tests, Homework, and Grading
For 191 students, 40% of the grade will be based on approximately 5 homework
assignments; 34% on a midterm test and a final noncumulative test (each
requiring about 80 minutes but answered overnight); and 26% on 2 staged
Lisp exercises (oriented towards "self-awareness").
For 291 students, the approx. 5 written homeworks will count as 35% of the
grade, the two tests as 24%, the 2 Lisp exercises as 16%, and the
project as 25%.
The homework assignments will ask you to answer questions and write
essays about the readings. These readings include several required books
(as follows) and a number of technical articles. (In addition, 291 students
will probably be looking at some further scientific literature to help
with the Gridworld consciousness project -- see below.)
Required books
Consciousness is often regarded as being of two sorts: access
consciousness -- the ability introspect and answer questions
about oneself -- and phenomenal consciousness -- our
subjective experience of colors, sounds, smells, pains, joy, etc.
For access consciousness, one can see rather readily how to achieve
that in machines, at least in principle; it's essentially an AI
question. Whether machines could also have phenomenal consciousness
is a more difficult question, though it's also surely of interest
in AI and robotics.
Hundreds of books and thousands of articles have been written, and
continue to be written, about consciousness, but there is very little
worthwhile literature on consciousness from an AI perspective.
The most useful book taking that perspective remains the first in
the following list.
Drew McDermott, Mind and Mechanism, MIT Press, 2001 (paperback).
[Though the book is out of print, it is still available on Amazon, incl.
Kindle. We'll be looking particularly at chapters 2, 3, 4, and 6, wich
respectively provide glimpses of AI, a "computational theory of
consciousness", a defense of that theory, and thoughts on robots,
people, and morality.]
Susan Blackmore, Consciousness: A Very Short Introduction,
Oxford U.P., second edition.
[An easily readable, entertaining yet quite comprehensive overview
of the riddle of consciousness and the various schools of thought
concerning it. One can question the conclusion that the "stream of
consciousness", and much else in our self-perception, is an illusion,
but the book is an excellent place to start learning about philosophical
and psychological views of consciousness.]
Bernard J. Baars, In the Theater of Consciousness: The Workspace
of the Mind, Oxford U.P., 1997.
[A delightfully written book about consciousness, demonstrating many
interesting points directly in the text. Baars is a cognitive scientist
with a thorough knowledge of neuroscience, but at the same time with
a high-level perspective on brain functioning, of the sort that is
very congenial to an AI perspective. He proposes a blackboard-like
model, with a workspace serving as "the immediate memory system in
which we talk to ourselves, visualize places and people, and plan
actions." The global availability and global influence of the workspace
contents make it the "theater of consciousness", though this metaphor
is a bit overworked in the book, and we'll primarily pick out parts
mental imagery, the inner voice, animal consciousness, etc., that are
touched on rather briefly in Blackmore. There's also a 2007 book with
broader scope by Baars & Gage -- see the link to further books
below.]
Jean-Marc Fellous & Michael Arbib (eds.), Who Needs Emotions:
The Brain Meets the Robot, Oxford U.P., 2004.
[Articles on emotion in humans and robots from various perspectives --
neurochemistry, functional anatomy of the brain, AI, robotics, etc.
We will probably cover chapters 7, 5, 10, and 12, in that order.
An interesting supplement is a 2018 interview with Joseph Ledoux,
https://www.youtube.com/watch?v=t1sbMWOzK84 .]
Time permitting, additional readings may be chosen from the further
books list (see link below). This might include some chapters of
Marvin Minsky, The Emotion Machine (Amazon.com 2006),
and possibly from Douglas Hofstadter, I Am a Strange Loop
(Basic Books 2007), and/or Ray Kurzweil, How to Create a Mind
(Viking 2012), and/or Murray Shanahan, Embodiment and the Inner
Life: Cognition and Consciousness in the Space of Possible Minds (Oxford
U. P. 2010).
Links to further books and articles
The following are links to some
further books from which some readings may be selected (for you or
by you) and many articles of related interest.
Writing guidelines
You will be asked to answer questions in writing, and to write essays
of various lengths.
One kind of place to look for examples of the kind of writing to strive
for is in book reviews, in refereed journals, of some of the readings
for this course.
Typically, essays will be graded on 3 dimensions, roughly as follows:
- work, 40%: How much effort in terms of careful reading of the relevant
materials, thinking, analysis and writing has gone into the preparation of
of the essay?
- coherence, 40%: How well organized and defended are the thesis or theses
of the essay? The thesis (or perhaps 2 or 3 theses) should generally be
clearly set forth at the beginning, and convincingly developed and argued
for in the rest of the essay, and there should be brief concluding remarks.
Use direct, clear language, illustrate your points with examples wherever
possible, and be sure to use proper citation style. (Also, you are
reminded of the academic honesty guidelines, stated below, concerning
all materials taken from others, which must be clearly identified and
acknowledged.
For more detailed guidelines about good essay style/organization, see
the notes by Mark Andrews
(written at Cornell U. in 2001; he is now a researcher in the Psychology
Dept. at University College London); and the
notes by John Messerly
(formerly at U.Texas at Austin and Austin Community College, now seemingly retired).
- Originality, 20%: This refers not so much to coming up with entirely novel
ideas, claims, or evidence (though that is welcome!), but rather to
the development of your own, personal critical/ constructive/ imaginative
perspective on the themes under discussion.
Project: Implementing a limited example of (self)awareness
As noted, 25% of the work in CSC 291 will be devoted to a semester-long
project, carried out in teams of about 1-3, developing "Gridworld"
agents that show or simulate some degree of perceptual awareness,
activity of some sort, emotional response, and self-awareness. Such
a system should contain a simple model of the system itself, i.e.,
a representation that includes a symbol for itself and symbolic
representations of other relevant entities in the "world" and how
they relate to the agent and each other. It should also be able to
do some simple inferencing. The existing Lisp-based "Gridworld"
environment for creating agents and simulated worlds will be made
available and also discussed in a lecture (and perhaps a recitation).
THE CONSTRUCTED AGENTS SHOULD SHOW SOME LIMITED QUESTION-ANSWERING
ABILITY CONCERNING THEIR CURRENT PERCEPTIONS AND SITUATION, AND THEIR
OWN GENERAL PROPERTIES, SINCE THIS IS THE MOST NATURAL WAY TO PROVIDE
OVERT EVIDENCE OF CONSCIOUSNESS. ANOTHER HIGHLY RECOMMENDED FEATURE IS
TO INCLUDE SOME EMOTIVE OUTPUTS, SUCH AS "MM... THE PIZZA IS DELICIOUS!"
WHEN SOME REWARD IS GAINED, OR "I'M SO HUNGRY" IF APPROPRIATE." (More
ambitious goals would include being able to say something about past
events, or handling more than one mobile agent, or a changing world. There
is a multiple-agent version pf the Gridworld, and "exogenous" changes
to the world, i.e., ones not caused by the agent, could be implemented
by expanding the "go!" function, which causes the agent to make new
plans and perform a further action, so that this also alters the
world incrementally.)
The prior Lisp exercises assigned to both 191 and 291 students
will be designed to resemble, in a very limited way, what can be done in
Gridworld.
Each team in the 291 class will attempt to go as far as possible
in demonstrating the above sorts of capabilities. Teams should form,
with the help of the grad TA, by about Spring Break, and project proposals
by those teams should be provided to the instructor and the TAs no later
than Thu. Mar 26, so that they can receive feedback about content and
feasibility. The proposals should be about 1 printed single-spaced page
in length. (There is no prescribed organizaton or grade for the proposal
-- it should just lay out your ideas and approach to achieving the above
project goals, so that you can get feedback before getting too deeply into
an attempted implementation.)
Final project presentations will be given near the end of the semester,
and reports on the projects will be submitted. Besides a system description
and examples of behavior, the final report should state conclusions about
what has been learned, about the implications for consciousness, and
connections between the project and topics/readings in the course.
Attendance and tests
Attendance via Zoom is essential for keeping up with the material and doing
well on the tests.
The instructor should be informed of unavoidable
absences in advance (whenever feasible).
Essays and reports must be handed in in time.
Exceptions will be made only under dire circumstances.
Academic Honesty
Student conduct in CSC 191/291 is governed by the College
Academic Honesty Policy, the
Undergraduate Laboratory Policies
of the Computer Science Department, and the
Acceptable Use Policy of Academic Technology Services (see
especially sec. III of the full pdf document).
As in all intellectual endeavors, proper attribution of work is crucial
in the essays and project reports you submit. 291 students also need
to be sure to provide proper references to the software and documentation
they acquire for the project (naturally, describing any additions or
modifications they made). In your homework answers, essays and reports,
provide citations in standard form for any text and ideas taken
from other sources. Use of unattributed material is plagiarism.
Last Change:
19 April 2022 / Len Schubert