Artificial Intelligence World Group University -
- Where all is gathered under one roof for research.
(TODO break into a tab based access page)
This is a new listing of Articles, thoughts People, and places for Artificial Intelligence,
with an emphasis on Common Sense, Knowledge Representation, and Natural Language Understanding/Processing.
What is Artificial Intelligence?
Artificial Intelligence: the use of programs to enable machines to perform tasks which humans perform using their intelligence. (more defs)
Read about AI on Wikipedia
My main concern with AI is enabling a computer program to be able to process, learn and reason
with this common knowledge, to be able to excel in a wide range of non-specific information
tasks. Simple reading and question answering in a story, newspaper article, or novel.
It should be able to read a story on any general topic and be able to recall and apply other
world knowledge into the story correctly.
This goal unfortunatly requires a good procedure for the three subtasks, Natural Language Processing,
Knowledge Representation, and Common Sense/World Knowledge.
In theory, if the computer is endowed with good Knowledge Representation Database, including
a comprehensive common sense database, and is able to process and respond in plain-text english,
it will have the ability to process and reason with english texts.
The task of Common Sense Knowledge is probably the least studied area, though it is included
in many ways in knowledge representation task. There are two issues with this, one is how to
represent the knowledge gathered in a computer processible, and human accessible way. This
has not been accomplished yet to my satisfaction. I belive some sort of pure english
representation would be the best structure. The second task is actually collecting the
Common Sense knowledge. There are a couple of different groups who are doing this now.
Knowledge Gathering is usually done for expert systems and is limited in its breadth to a
limited domain. The two common sense projects I know about are
Open Mind Common Sense and Cycorp.
There is one other project type that is very helpful for common sense processing, the ontology systems,
such as WordNet.
I put forth that each of these has its downsides, and cannot stand alone as a common sense processor,
but in combination, and with an added piece of "Large Scale Word Processing" could possibly generate a
large scale usable database. By this I mean to statistically processes a large text database, of
over 1000 current general fiction books, and use them to extract common sense information.
Cyc Project - The Cyc project has been going for over 20 years now, with little to show for it.
"Doug Lenat's CYC project [Lenat and Guha 90], maintains that intelligence in systems stems
primarily from knowing a large number of simple facts, such as "Water flows downhill" or "If
someone shouts at you, they are probably angry with you". The intuition is that even simple
queries depend on understanding a large amount of context, common sense knowledge that remains
unstated, but is shared among most people with a common language and culture."
I have a few problems with Cyc, one major one appears to be that there does not seem to be a
stopping point to their rule-writing, nothing has congealed together into the Magic Number where
it can work by itself. Hmm Cyc seems to have cabbage, but not a chair? Odd.
Open Mind Common Sense OMCS - Again, this was a greatly interesting project, and showed a
couple of things, namely that thru the use of millions of users, data could quickly and decently be gathered,
even on such a topic as common sense. Unfortunatly, this hasnt been taken a step further into a
nice usable construction. This system is however, able to gather facts at an astounding rate, and
a good quality and variety of knowledge. There are 3000+ statements about chairs.
WordNet - Not intended to be a common sense system in itself, this ontology has
been greatly helpful in creating these systems, and is worth mentioning here. Its problems include
an upper level ontology that is vague and not based on common english phrases and words, and the
level of correctness varies across the dictionary. Overall though it is a great help, and allows simple
things like knowing a chair is a seat, and a couch is a seat, so a chair is like a couch. I use
this, and then add on other common sense uses gathered from novels. 6 Senses of the word "chair"
but low frequency counts, and many little-known used words are degrading to the system.
Common Sense reasoning is by nature centered around the way humans think and react. But these actions
can be broken down further into specific areas of understanding.
NLP and mechanical actions have been all discussed before, but the main need for reasoning is the
knowledge gathered and the way it is processed and dealt with.
I am looking at several categories here:
People - We interact most often with other people, every day we have a whole network of people we
interact with, how we interact, what is done what is said, and what relationship we have with these
people is of paramount importance when reasoning about them.
Places - We interact differently according to where we are. We wouldnt do something at work
that we may do at home, or at a university environment. Location is a necessary piece of knowledge
for adequate reasoning. This can be gathered specifically from GPS coordinates.
Time - All reasoning is context sensitive to time as well. This is easily gathered from an
Things - We interact daily with things, a huge database of object knowledge is necessary.
Learning - A system must be able to learn easily, and naturaly about all of the above items.
This is of utmost importance, just beneath the Reasoning Tasks. This is a general idea that can be
applied to all applications of AI, unlike the Reasoning Task.
Beginnings of Artificial Intelligence -
Artificial Intelligence: the use of programs to enable machines to perform tasks which humans
perform using their intelligence.
Common Sense: beliefs or propositions that in their opinion they consider would in most people's experience, knowledge held by people "in common". (more)
Common Sense in Humans 2001 - Good Overview of Common Sense
Erik T. Mueller MIT
Prospects for in-depth story understanding by computer 1999
Erik T. Mueller MIT
Commonsense Reasoning in and over Natural Language - 2004 (PDF) (Conceptnet)
Hugo Liu and
Push Singh -
MIT Media Lab - uses WordNet and
Open Mind Common Sense)
Teaching Machines about everyday life 2004 Singh, Barry, Liu
Collecting Commonsense Experiences - 2003
Hals Legacy - 2001 (HTML) - Doug Lenat of CycCorp
Can We Derive General World Knowledge from Texts? - 2002 -
Lenhart Schubert of University of Rochester
Commonsense-Based Interfaces 2000 Marvin Minsky
Draft of Chapter 6 of The Emotion Machine 2004 Marvin Minsky
The St. Thomas. Common Sense. Symposium:. Designing Architectures for Human-Level Intelligence. 2004
Push Singh, and Aaron Sloman
WHY PEOPLE THINK COMPUTERS CAN'T - Minsky
Out of Context: Computer Systems That Adapt To, and Learn From, Context 2000 -
Henry Lieberman and
Ted Selker MIT
Commonsense and Embodied Agents: A Panel Discussion 2001 - Stuart C. Shapiro, Eyal Amir, Henrik Grosskreutz, David Randell, and Mikhail Soutchanski
Knowledge Representation - Introductory Article - Stuart C. Shapiro
The Structures of the Common-Sense World 1995 (Naive Physics) - Barry Smith
Consciousness: Drinking from the Firehose of Experience. - Mooney, R. J. and Bunescu, R. 2005
Mining Knowledge from Text Using Information Extraction - Benjamin Kuipers. 2005
When is a Cognitive System Embodied?, Alexander Riegler
- This is an interesting article that goes along with my thoughts on embodiment, though
I go further to say it needs to be physically embodied.
Research Facilities and Universities
I am also interested in pursuing DARPA grants :
DARPA - Information Processing Technology Office (IPTO) (Cognitive Systems)
The Main project is
Cognitive Information Processing Technology (CIPT)
Over the years, DARPA has sponsored Text REtrieval Conference (
Message Understanding for Comprehension (
wiki), speech understanding evaluations, robotic competitions,
High Performance Knowledge Bases (
HPKB) and Rapid Knowledge Formation (
knowledge-base evaluations, and many more.
NSF Research Grants
I am most interested in Strong AI (Searle)
Strong AI is the actual development of high level AI that is concious/autonomous and operates at a human level intelligence
New Tools used in AI Research include Google Books, and Google Scholar
Search G Books for AI (173,000 results)
Search G Scholar for AI (551,000 results)
G Books: "common sense" AI
G Scholars: "common sense" AI
Google AI Military Sites
Formal Ontology, Common Sense and Cognitive Science
To Read articles:
Representing Knowledge Soup In Language and Logic
John F. Sowa
New Site: http://cogprints.org
ConceptNet Singh Push
ThoghtTreasure Erik Mueller
To Print and Read:
http://www.aaai.org - American Association for Artificial Intelligence
at Rochester NY
NEW CLASS: http://www.cs.rochester.edu/u/schubert/247-447/
KR - What is a KR?
Ontology and Information Systems
Ontology Development 101: A Guide to Creating Your First Ontology
Natalya F. Noy and Deborah L. McGuinness
Natural Language Processing with ThoughtTreasure
Text analysis and knowledge mining system
Who's Who? Identifying Concepts and Entities across Multiple Documents.
Taxonomies Put Content in Context by Russell Letson
NOTICE: ALSO LOOKUP SHORT-STORY UNDERSTANDING< IN THE FOR OF SAT AND GRADE SCHOOL READING COMPREHENSION WORKSHEETS!
Henry Lieberman's lecture on Applying Common Sense Reasoning in Interactive Applications (one hour, requires Realplayer).
http://web.media.mit.edu/~lieber/ his homepage
Story Understanding Resources:
Buffalo University in NY
The Plots of Children and Machines: The
Statistical and Symbolic Semantic Analysis of
1972 - Charniak - Toward A Model of Children's Story Comprehension
Frame Based Text Processing
login to acm to get this!
NSF Grants - sample one, do a search though for all AI common sense ones!
awards to: Bender, Walter Ashley, Kevin Schubert, Lenhart
Symposiums and Talks
aaai - The 2005 AAAI Spring Symposium Series
International Joint Conference on AI. 2005
INTERNATIONAL COMPUTATIONAL INTELLIGENCE SOCIETY
JOIN FREE, ADD BOTH OF THESE TO resume
July 9-13, 2005, Pittsburgh, Pennsylvania
Daydreaming machines book 300+ pages
NEw Mexico COnference
David Mark's Geography Spatail cognition
WordNet 1.7.1 online http://www.foxsurfer.com/wordnet/
Take this brain image
(TODO CREATE A DARPA IPTO RESEARCH PAGE)