Instructor: Samir Chopra
Office: 1214
Ingersoll Hall (Mailbox is in the CIS office, 2109 Ingersoll Hall).
Email: schopra@sci.brooklyn.cuny.edu
Phone: 718.951.4139
Class Meeting Times: Monday – Thursday 10:50 AM -12:05 PM (3214 Ingersoll Hall)
Office hours: Thursday, 12:10
PM – 1:10 PM
This course will attempt to deal with the following questions. What is artificial intelligence? Is it possible? If so, how is it realizable? What model(s) of mind does modern research on artificial intelligence assume? What are the philosophical arguments - good and bad - for and against artificial intelligence? And so on. The central question of course, is: can a machine think - or do X, where X is some uniquely human mental ability? If the answer is yes, what kind of machine can do X? To answer this question, we need to unpack the computational model of mind – what is a Turing machine and why is it sometimes held that the brain is a Turing machine? What is the neural network or neo-connectionist model and why is it said to be opposed to the computational model of mind? Does it offer a more promising avenue of research for AI practitioners? We will also look at arguments against AI – those that think that AI is impossible in principle, and those that think that AI is impossible in practice. What is the Chinese Room argument and why is it relevant to artificial intelligence? We will also discuss one of the most famous arguments against artificial intelligence – the so-called ‘qualia’ argument. Most intuitive (‘layman’) arguments against artificial intelligence are some variant of this particular argument. If artificial intelligence is possible in principle, what problems does it face? What is the so-called frame problem? What is the tractability problem? What are the arguments in the technical computer science community about how to realize artificial intelligence?
Course Requirements:
There will be one mid-term examination, two paper
assignments - one before the mid-term and one before the final exam - and one
final exam. Each week, readings will be assigned. All readings listed in the
reading list are required in that they can, and will be, the subject of paper
assignments and examination questions.
Note: Paper assignments must be
handed in on time i.e., on the date and time specified. Late paper submissions
will receive an automatic grade of zero. I will be willing to look over a late
submission to give you feedback, but I will not assign you a grade for it.
The semester grade will be determined as follows:
Paper assignments: 15 % of the final grade (each).
Midterm exam: 30 % of the final grade.
Final exam: 40 % of the final grade.
Class Mailing List:
Go to www.sci.brooklyn.cuny.edu/cis/majordomo/?philai_fall03, pick
the subscribe (default) option (the list name is philai_fall03) and
enter your email address. Alternatively, send a message to majordomo@sci.brooklyn.cuny.edu with a
blank subject line and subscribe philai_fall03@sci.brooklyn.cuny.edu as the
body of the message (caution: if you are using a web-based email service, make
sure it does not insert HTML in the body of the mail - using the web page above
is safer). Subscription to the class mailing list is mandatory – consider this
your first homework assignment.
Required Textbook:
Jack Copeland, Artificial Intelligence: A
Philosophical Introduction, Blackwell Publishers, Oxford, 1993. ISBN:
0-631-18385-X
Copeland’s book has a slightly dated feel to it in the
sections on computer science and AI research. However, it does a magnificent
job in the philosophical sections in making things clear and accessible and
provides a good road through the central debates. Where and when it mentions
these, we will read the literature as necessary. Since the textbook was ordered
through the CIS Department, the bookstore has it displayed in the CIS 32.1 bin.
Readings:
It’s a truism but one worth paying attention to again,
that the more you read this semester, the more you will get out of this class.
The topics covered in this class are not simple, and require considerable attention
and diligence in reading the material indicated. Fortunately, these labors pay
rich dividends. To do justice to the class, yourself and your fellow students,
it behooves you to work hard at the readings. If you find the vocabulary
unfamiliar, work with a dictionary. If you find material hard going, don’t
worry, it is. Work through the readings closely, and, if need be, more than
once. Your classroom experience – and mine in turn - will be immeasurably
enhanced and enriched if you come to each lecture having worked through the
indicated readings. Conversely, the classroom experience – everyone’s, not just
yours - will be considerably diminished if you insist on coming to class
unprepared.
Bullet listed readings below are included in the course
packet. You will notice as you work through the readings that the location of
the readings in terms of coverage of the material is not exact – some of the
readings will apply to more than one portion of the semester (for example,
Crane’s chapters refer to the Chinese Room argument and so on). These parts can
be revisited and other readings brought forward – we will take it as it
comes.
Chapter 1: History of AI: a brief examination of the
history of computing relevant to AI. The kickoff papers here will be:
Chapter 2: This is by far the most dated chapter in
Copeland’s book, but is still good value. You can chase down information on the
various projects mentioned in this chapter on the Internet. For an example of
one of modern AI’s achievements check out IBM’s Deep Blue website: http://www.research.ibm.com/deepblue/.
For a discussion following Deep-Blue’s over Kasparov, look at
slate.msn.com/id/3650/, in which Daniel Dennett and Hubert Dreyfus spar over AI
and at http://www.pbs.org/newshour/bb/entertainment/jan-june97/big_blue_5-12.html.
Another opinion is ftp://ftp.cs.yale.edu/pub/mcdermott/papers/deepblue.txt.
There is plenty more written on Deep Blue – you can find a fair amount of
material on the Internet.
Chapter 3: The basic disconnection between thought and
consciousness, the Turing test and responses to it. An objection to the Turing
test may be found in:
Chapter 4: This covers the fundamentals of computing and the
computational model of mind. It will be tempting to think “I’m just going to
let my eyes glaze over on this chapter” but that would be a mistake. To
understand computing, its basic assumptions and their significance for AI, this
chapter is essential. The supplemental readings for this chapter are as
follows:
Chapter 5: The problems with AI – the very idea! Various
essays on the infamous frame problem may be found in:
An enduring critique of AI is to be found in:
Drew McDermott launches an attack on his fellow computer
scientists in:
Steve Hanks and Drew McDermott offer a slightly more
technical - logic wise - discussion in:
·
Steve Hanks and Drew McDermott, ‘Default Reasoning,
Nonmonotonic Logics, and the Frame Problem’, In Proceedings of the National
Conference on Artificial Intelligence, Morgan-Kaufmann, pages 328-333,
1986. This essay is of particular
interest in making the strong claim that non-monotonic formalisms are unable to
capture commonsense reasoning. Much, much work has been done since then in commonsense
reasoning and reasoning about action. An Internet search will be
particularly useful in tracking down some of the latest work (http://www.ucl.ac.uk/~uczcrsm/researchers.html
might be a good place to get started).
Check out Murray Shanahan’s Solving the Frame Problem (MIT Press,
Cambridge, 1997) for a book-length logic-based treatment of the frame problem.
John McCarthy mounts some defenses
of AI in (what else?):
·
John McCarthy, Defending AI Research (CSLI
Publications, Stanford, 1996). We will
read essays #s 5,8,9,10,12,13 (each piece is fairly short).
You might want to track down Roger Penrose’s The
Emperor’s New Mind (Oxford, 1989, reprinted 2002) and Shadows of the
Mind, (Oxford, 1996). These books were written as popular science books,
but have a fair amount of detail on Turing machines, Goedel’s theorem,
technical notions of computability and quantum mechanics. An online symposium
on Penrose may be found at: http://psyche.cs.monash.edu.au/psyche-index-v2.html.
I do not think that we will cover the Penrose argument because of the technical
details, but time permitting, we might.
Allied to the Penrose argument is the Lucas argument
against AI (again, its not clear at this time that we will have the time to
cover this):
·
J.R Lucas, ‘Minds, Machines and Gödel’, Philosophy,
36:112-127. An online version can be found at:
http://cogprints.ecs.soton.ac.uk/archive/00000356/00/lucas.html
A simple search on the Internet will also bring you to many,
many arguments against these two positions (look at the section in David
Chalmer’s bibliography listed below).
Chapter 6: A full discussion of the Chinese Room argument
will be found in:
The Chinese room and the responses to it are a little cottage
industry all by itself. Look at Chalmer’s bibliography below to chase down
references and to check out all the various shadings of the debate. A book
length treatment of the Chinese Room argument can be found in John Preston’s Views
into the Chinese Room, (Oxford University Press, Nov 2002).
Chapter 7: (Free Will): We will cover this only if time
permits – more interesting material elsewhere is available for discussion.
Chapter 8: (Consciousness): The nasty business of qualia:
the basic argument against AI on the basis of qualia is easily understood.
Supplemental reading for this chapter is:
·
Daniel Dennett, ‘Quining Qualia’, in Alvin Goldman ed., Readings
in Philosophy and Cognitive Science, MIT Press, Cambridge, pp 381-414,
1993.
Chapter 9 (The Brain just is a computer!): Copeland
unpacks the strong symbol system hypothesis and its problems. Supplementary
readings are the same as those indicated for Chapter 4 i.e., Newell and Simon
in Mind Design I and the extended treatments by Block and Pylyshyn.
Chapter 10: (Connectionism, neural nets, PDP, cybernetics): This chapter explains connectionism and the neural network model as a competitor to the computational model of mind. Supplemental reading for this chapter is as follows:
·
David Rumelhart, ‘The Architecture of Mind: A
connectionist approach’, in John Haugeland, Mind Design II, MIT Press,
Cambridge 1997.
·
Paul Smolensky, ‘Connectionist Modeling: Neural
Computation/Mental Connections’, in John Haugeland, Mind Design II, MIT
Press, Cambridge 1997.
·
Jay Rosenberg, ‘Connectionism and Cognition’, in John
Haugeland, Mind Design II, MIT Press, Cambridge 1997.
·
J. Fodor and Z. W. Pylyshyn Z., ‘Connectionism and
Cognitive Architecture: A critical analysis’, in John Haugeland, Mind Design
II, MIT Press, Cambridge 1997.
·
Jack Copeland, ‘The Church-Turing Thesis’, in Stanford
Encyclopedia of Philosophy.
A history of cybernetics (the movement that led to
connectionist models of the mind) and its philosophical foundations can be
found in Jean-Pierre Dupuy’s Mechanization of Mind, Princeton University
Press, 2000.
Time permitting, we will consider some
anti-representational polemics:
Classroom Participation:
Participation in class discussion is of paramount importance. Your understanding of topics will often be best displayed by your asking intelligent, well reasoned questions, raising valid objections to arguments and displaying an awareness of the issues raised in any particular reading. Nothing you say in class will cause your grade to be lowered. Ask questions on my lectures and the readings. You will aid your understanding of the material, and more often than not, help me as well. Each class discussion will build upon previous sessions. Absence from a class means that you will not profit from the class discussion and that your ability to answer exam questions and write relevant papers will be diminished.
Web resources and other miscellany:
Our reading list above is neither complete nor
exhaustive. (It makes no pretensions in this regard). Please use the following
links – the collection below is not exhaustive either! - to look for additional
reading for topics covered in class, as research resources for your paper and
examination preparation and of course, to further pursue your readings if this
area interests you. A good strategy for broadening your readings on AI would be
to look at the websites of the authors mentioned in the reading list to look at
other writings of theirs – and to chase down links they provide to other
writers and so on.
Jack Copeland provides an overview of AI in www.cs.usfca.edu/www.AlanTuring.net/turing_archive/pages/Reference%20Articles/What%20is%20AI.html.
Rodney Brook’s website at MIT (www.ai.mit.edu/people/brooks/index.shtml)
has lots of interesting information on Brook’s alternative paradigm for AI.
Look for information on Cog and Kismet. Brooks also makes an appearance in the
movie ‘Fast, Cheap and Out-of-Control’. Check it out:
http://us.imdb.com/title/tt0119107/.
John Sutton has an excellent page
(http://www.phil.mq.edu.au/staff/jsutton/363reading.html) for his class on
Philosophy and Cognitive Science; you might find readings here that are more to
your taste on some of the topics that we cover in class. I might make some
readings from this collection available to the class during the semester.
Bibliographies may be found at www.earlham.edu/~peters/courses/mm/mmlinks.htm
and www.u.arizona.edu/~chalmers/biblio/4.html