An Artificial Intelligence System
Your task is to describe and demonstrate an aritificial
intelligence system. Remember that the project is also to be done in
pairs; if you didn't have a partner for previous assignments, you are
encouraged to have one this time.
Your are asked to write a 2000-3000 word description of the system
and prepare a 15-minute presentation. Follow the hints and
requirements below.
Submission
In class on April 8: submit a one-page overview description (around
500 words) of the AI system you are investigating. Make sure to
include relevant references (including URLs to the system).
By 4pm on May 7: submit a printout of you complete description of
the system and slides of your presentation. Email cse352ATcsDOTsunysbDOTedu before that
a .zip file that contains all your documents.
Printouts are preferably printed double sided and or 2 pages per
sheet.
Grading
This project is worth 20% of the course grade. Exceptionally
well-thought-out and well-written homeworks will receive appropriate
extra credit. Partial solutions get partial credit.
You will receive extra credit if you describe good ways to improve
the functionality, performance, or other aspects of the system and
possibly implement some improvements.
Hints and Requirements
- Never start by writing such an AI system from scratch yourself! :)
Instead, start by thinking what kind of systems you would like to
see/have/learn/use and then search the Web for such systems. If the
system you finally chose happens to be the same or related to the one
you described in Assignment 1, note that at the end of your writeup.
- The system you pick should preferably involve uncertainty
reasoning (for example involving probability) or learning (for example
using neural networks). This is not required, but it will help you
better understand the materials under discussion in class.
- The system you pick can be in any programming language running
on any platform, as long as it is in a language that you know and on a
platform that you can run. This is so that you can know some more
interesting details than what is usually in the advertisement and so
that you can run the system.
- Your 2000-3000 word description should have the following components:
- An overview description: what the system does, what are its
applications, what techniques it uses, who developed it, who use it,
and anything else that you think is important or interesting for the
system you pick.
- A description of important functionalities of the system: what
they are, and how to use them. Usage, and possibly demo, usually show
the functionalities well. You should try to both summarize over the
kinds of usages and give results of specific runs on examples.
- A description of the architecture of the system and the
techniques used in the systems. What searching and problem solving,
knowledge representation and inferencing, uncertainty reasoning, or
learning techniques are used? What is the structure of the system.
What language is it implemented in? Include any other important or
interesting, general or specific, aspects of the implementation that
you find as well.
- As bonus, a description of your suggested improvements and
implemented features, if any. Make sure you have well-documented
code, show runs with your improved features, and give instructions for
installation and testing.
Everything in CS and AI should have "why" (why a system is
relevant), "what" (what it does), "how" (how it does what it does).
So, the first item above is like the "why", with very brief summaries
of "what", "how", and other important aspects. The second and third
items are like the "what" and "how", respectively. The last item is
what is exciting. :)
I am sure you could cut and paste a lot of things from the
documentation of system you pick, but make sure you understand and
distill them, and really try the system. You need to demonstrate that
you know the actual system, not just the system documentation, both
through your writing and your presentation.
- Your 15-minute presentation should have the same materials, but
presented succinctly with summaries, highlights, pictures, etc. You
may include a real demo of the system if it is possible, i.e., it can
be set up and runs on the machine in the teaching room, and if it is
appropriate, i.e., it shows the system better and shows your knowledge
of the system better than the slides do.