Here you will find a collection of writings about a variety of topics, published, unpublished, or published in relatively obscure places. I have organized them broadly by theme and note that this section is under active development.
I have managaged to be part of a range of research laboratories and intimately involved in the creation and gestation of a few of those. In addition, part of my scientific research (especially my PhD work) was about creativity and its foundations. As a result, I have ideas about what works and what doesn't when making contexts for creativity. I've put some of these ideas into writing, mostly based on personal experience enriched in some cases with the stories of others.
Why the Media Lab Works in IBM Systems Journal, Fall 2000. [pdf]
Arts & Engineering: Strange Bedfellows or Separated at Birth online presentation for University of California at Irvine (as part of a proposal for a major collaboration between their Schools of Arts and Engineering).
How To Do Corporate Research [pdf], an outline for a corporate research laboratory which emphasizes the special pressures which need to be addressed to make corporate research laboratory fresh, relevant, and excellent.
Much of my technical work has been focused on making large knowledge bases robust, efficient, and ubiquitous. This work has been about creating the capabilities that intelligent agents do and will need to be the everyday miracle we imagine they will become. The work started in the early 1980s with a reimplementation of Greiner & Lenat's RLL, which was called ARLO. Subsequently, I ended up implementing another version (ARLOtje) as a teaching tool. At the Media Lab, I implemented a first large-scale (>100,000 frame) knowledge base called Framer which was used for several of my student's theses. This led to the FramerD knowledge base infrastructure with is still an active project (www.framerd.org).
BRICO: Building an Inter-Lingual Ontology in IBM Systems Journal, Fall 2000. [pdf]
FramerD: Representing Knowledge in the Large, in IBM Systems Journal, Fall 1996. [pdf]
"Framer: A Persistent, Portable, Representation Library", Proceedings ECAI-94, 1994
"Supporting Knowledge Representation in the Large", Workshop on AI in Service and Support Applications, AAAI-93, 1993
In Godel, Escher, Bach, Douglas Hofstadter made the partially true and quite contentious assertion that "all AI advances are language advances". But I have repeatedly learned (and in my PhD thesis, proven) that the right language and representation can make impossible problems trivial. (Contrariwise, some representations can make trivial problems seem impossible, as expressed in this ditty and far too many patent filings).
In addition, my early intellectual years were spent as part of Gerry Sussman's group at MIT, where thinking about languages and language design were part of the atmosphere. I've implemented numerous languages over the years and have forgotten many of them, but I have written about some of that work, which I present here.
Soft Objects, describes an object system based on functional extension and augmentation rather than abstract schematization.
Utildocs, describes the many Scheme utilities I implemented as part of my PhD thesis. One of them, PRINTOUT, survives into FramerD and was originally inspired by Interlisp's formatted output facilities.
I have long been interested in the nature of creativity and the use of computational models to describe it. In fact, I did an experiment in high-school with a program which created simple melodies based on equally simple rules. Primarily, though, my interest has focused on the role that representation plays in discovery and that the invention and evolution of representations plays in supporting discovery. Following my ealry mentor Thomas Kuhn, I am also a critic of many reconstructions of past scientific discoveries with computational methods, since they often end up giving the programs the very representations where the crux of the innovation lies.
Machine Discovery, in Machine Learning, edited by R. Forsyth, Chapman and Hall 1989
"Discovery Systems", Proceedings of ECAI-86, 1986
"Cyrano-3: An Experiment in Representational Invention", Workshop on Machine Learning, International Conference on Machine Learning, 1992
"Too Many Ideas, Just One Word: A Review of Margaret Boden's `The Creative Mind'", AI Journal, Volume 79, Number 1, December 1995.
PhD Thesis
One of the significant results of my work on creativity and discovery was that the evolution of representations was key to creativity and that thos revolutions were often tied to concrete instances and examples. In concert with Kuhn's focus on exemplars and paradigms as the core of scientific theories, this led me to think about representations where analogy rather than a priori structures shaped representation. In a certain sense, this broke with much of what I had done before in thinking about representation languages.
Making Clouds from Cement: Building Abstractions from Concrete Examples, in Vision and Language across the Pacific, edited by Y. Wilks and H. Okada, Ablex 1996. [ps]
Matching Texts to Extract Information [ps]
"Analogy in the Large", submitted to IJCAI-95. [ps]
"Why No Representation is Better than One (and Many are Better still)", unpublished talk at Marvin Minsky's Festschrift, August 1994.
"A Uniform Memory-based Representation for Visual Languages", Proceedings of ECAI-92, with Anil Chakravarthy and Louis Weitzman, 1992.
"A Model of Poetic Comprehension," Proceedings AAAI-96, AAAI Press 1996
When I came to the Media Lab, it became clear that my previous work, on programs which devised representations to rediscover simple arithmetic, was unlikely to charm the sponsors. In addition, Nicholas Negroponte's recurring advice to new faculty was "do something different". As a result, I began to look at using language and knowledge in information retrieval. This involved creating robust parsers, looking at how language can be used, and thinking about the roles that could be served by machine understanding and the capabilities neccessary to support it.
Hands For The Mind: How We Think With Models, book, in preparation. [html]
Do Agents Need Understanding? in IEEE Expert, February 1997.
"Using Semantic Knowledge for Information Retrieval", Proceedings SIGIR-95, 1995. [ps]
"Multi-scale Parsing Using Optimizing Finite State Machines", unpublished technical report 1992
"Using Grammatical Relations for Document Analysis and Retrieval", 1994. [ps]
Some things defy categorization and some of my writings don't clump into the bulging packets gather together above. Here are some other things I've written that you might find interesting. Enjoy.
Being On Time: Parsimonious Approaches to Responsive Action [PDF] began as my PhD area examination and makes some interesting points about representation and interaction with the world. It's chief point was that work on situated action included representation whether you liked it or not.
The Rise and Fall of an Evangelical Atheist describes the early phases of my own spiritual journey, telling the story from a cognitive developmental standpoint. It is a little self-disclosing and I appreciate your reading it with an open mind.