Research of Marc Pickett I

Creation of a machine that we can call intelligent has been the stated goal of the field of Artificial Intelligence for the past 50 years. However, most traditional applications in Artificial Intelligence and Robotics rely heavily on human knowledge engineering. This means that the programs depend on human programmers to interpret their data and develop their representations, so the programs tend to fail when faced with situations that the human programmer did not anticipate. If a truly intelligent machine is to ever emerge it will need to be capable of handling such novel situations. Therefore, this project aims at developing algorithms for automatic development of representations, characterizations, and abstractions of raw uninterpreted data.

Learning from raw data can be challenging. Consider a mobile robot with an array of sonar sensors. When plotted over time, the raw data that the robot receives from these sensors looks like a set of "squiggly lines" that are almost unintelligible to a person seeing them for the first time. At this level, concepts that humans sometimes take for granted such as "Obstacle", "Doorway", and "Trash Can" are highly abstract entities. "The Squiggly Line Problem" is that of autonomously developing a description of the world that includes such abstractions starting from raw uninterpreted sensor data. The goal of this research is to develop a general purpose representation schema and abstraction algorithm that will have applicability in solving the Squiggly Line Problem.


(An article about applying my work to The Netflix Prize appeared in the front page of the Sunday edition of The Baltimore Sun on December 10th, 2006 (and later in the L.A. Times). Here's my blurb about it.)

Publications

[1] M. Pickett, D. Miner, and T. Oates. Essential Phenomena of General Intelligence. In Proceedings of The First Conference on Artificial General Intelligence, 2008.
[ bib | .pdf ]
[2] M. Pickett, D. Miner. Representation Change in The Marchitecture. In Working Notes of the AAAI Fall Symposium on Representation Change, 2007.
[ bib | .pdf ]
[3] M. Pickett, D. Miner, and T. Oates. A Gauntlet for Evaluating Cognitive Architectures. In Working Notes of the AAAI Workshop on Evaluating Architectures for Intelligence, 2007.
[ bib | .pdf ]
[4] M. Pickett and T. Oates. The Marchitecture: A Cognitive Architecture for a Robot Baby. In Proceedings of the 22nd AAAI Conference on Artificial Intelligence (AAAI-07), 2007. Student abstract and poster.
[ bib | .pdf ]
[5] M. Pickett. The Übercruncher: Concept Formation by Analogy Discovery. In Proceedings of the 22nd AAAI Conference on Artificial Intelligence (AAAI-07), Doctoral Consortium, 2007.
[ bib | .pdf ]
[6] M. Pickett. Using Analogy Discovery to Create Abstractions. In proceedings of the 7th International Symposium on Abstraction, Reformulation and Approximation (SARA 2007), Lecture Notes in Artificial Intelligence. Springer Verlag, 2007.
[ bib | .pdf ]
[7] G. Chaddock, M. Pickett, T. Armstrong, and T. Oates. Models of Strategic Deficiency and Poker. In Working Notes of the AAAI Workshop on Plan, Activity, and Intent Recognition (PAIR), 2007.
[ bib | .pdf ]
[8] M. Pickett and T. Oates. The Cruncher: Automatic Concept Formation using Minimum Description Length. In proceedings of the 6th International Symposium on Abstraction, Reformulation and Approximation (SARA 2005), Lecture Notes in Artificial Intelligence. Springer Verlag, 2005.
[ bib | .pdf ]
[9] N. Berry, T. Ko, T. Moy, M. Pickett, J. Smrcka, J. Turnley, and B. Wu. Computational Social Dynamic Modeling of Group Recruitment. Sandia Report SAND2003-8754, Sandia National Laboratories, 2003.
[ bib | .pdf ]
[10] M. Pickett. Research Summary. In proceedings of the 5th International Symposium on Abstraction, Reformulation and Approximation (SARA 2002), Lecture Notes in Artificial Intelligence. Springer Verlag, 2002.
[ bib | .pdf ]
[11] M. Pickett and A. Barto. Policyblocks: An Algorithm for Creating Useful Macro-Actions in Reinforcement Learning. In Proceedings of the International Conference on Machine Learning, 2002.
[ bib | .pdf ]

Contact
home Last modified: Wed May 16 12:13:21 EDT 2007