Most project themes should be chosen from the following topics,
or in consultation with the instructor.
Miscellaneous pointers or comments are included with some of the subtopics;
these should not be taken too seriously -- they are just things found
"by accident" or by cursory googling. More will probably be added
incrementally. When you choose a topic, you'll research it much more
thoroughly! One goal will be to find the best system that can be obtained
and demonstrated to the class, in the chosen subfield.
Understanding stories and descriptions in NL
Engaging in NL dialogue
Representing and reasoning with common-sense knowledge
The goal is eventually to match humans in making common-sense inferences
about the kinds of things
(people, places, objects, beliefs, events, etc.)
everyone knows about. See e.g., Common Sense Problem Page,
http://www-formal.stanford.edu/leora/cs/
Playing intellectually challenging games
Uncertain inference
Forming and reasoning about goals and plans
(see last 2003 issue of JAIR -- special issue on planning, based on papers
from IPC 200.)
Automated mathematical reasoning and theorem proving
David M. Kaiser, Playing games with theorem provers,
http://www-users.cs.umn.edu/~dliang/CSci8980/Projects/TheoremProvers.htm
Selmer Bringsjord et al. at Rensselaer claim to have a system MARMML that
solves deductive, inductive,
creative, and visual inference problems:
http://www.cogsci.rpi.edu/marmml/MARMML.REAL.RPI.v3.ppt
Autonomous discovery, concept formation
The knowledge acquisition bottleneck -- acquiring propositional
knowledge about the world
Computer assisted learning/tutoring/instruction/education
One random item: D.G. Moursund, Brief Introduction to Educational
Implications of Artificial Intelligence,
2003 (an on-line book, it seems)
http://www.uoregon.edu/~moursund/AIBook/ai-contents.htm
Computers and law -- esp. computer judges
One potentially interesting article: H.J. van den Herik, The Future
World of AI,
http://www.cs.unimaas.nl/~herik/presentations/future%20of%20AI.pdf.
Music composition
See papers on EMI; also Robert Rowe, "The future of AI and music"
(which also refers to a 2001 book by
Rowe, "Machine Musicianship"),
http://students.washington.edu/shrecg/aimusic/future.html
Computer-generated art
Bioinformatics
Try the following (though it timed out when I tried):
http://www.doc.ic.ac.uk/~sgc/slides/
Swarm intelligence, multi-agent problem solving
AI through ALife (artificial evolution, genetic programming, ...)
The Golem project at Brandeis, by Hod Lipson and Jordan Pollack, is
tangentially relevant (though there's
no real concern with intelligence
here; but there's a good "Further reading" list at the end):
http://www.demo.cs.brandeis.edu/golem/.
One person, Karl Tiedemann, suggests that neural nets developed through
genetic programming are the way
to achieve AI:
http://www.karlt.com/blog_archives/000018.html
AI through quantum computers
e.g., see Subhash Kak's web page,
http://www.ee.lsu.edu/kak/home.html.