|
|
|
|
|
|
|
The Genesis Group is dedicated to the proposition that the time is ripe to answer fundamental questions about the computations that account for human intelligence. Members of the group believe that if we're to understand the nature of human intelligence, we have to understand the contributions of our vision, language, motor, and faculty-connection mechanisms. We further believe we have to understand how those faculties make it possible to understand the physical world and how that understanding provides a foundation for abstract thinking and learning.
The roots of the Genesis Group lie in the thinking that led to the thoughts of Robert C. Berwick, Thomas F. Knight, Jr., Gerald Jay Sussman, Shimon Ullman, Patrick Henry Winston, and Kenneth Yip, a group that styled itself as The Human Intelligence Enterprise.
The first major step toward understanding the role of vision and language is to build a testbed in which vision and language work together to understand a projection of the physical world that is simple, yet features paths, agents, causes, both language and visual inputs and outputs, complex state changes, and support for abstract reasoning in abstract worlds. Accordingly, we have been developing the Bridge System, which uses typed English input to drive an imagination system. Given a sentence such as "The fast hawk flew from the top of a tall tree via a pink rock to the top of a yellow pole and then dove on a brown rabbit," the Bridge System conceives and displays an imaged sequence of motions in blob world.
Eventually, we expect to exploit both symbolic and nonsymbolic representations in the Bridge System, but in our initial work we have concentrated on the symbolic side, choosing to use a representation conceived by Jackendoff to describe the movement of animals and artifacts along trajectories and a representation conceived by Borchardt for describing state changes in terms of a vocabulary of qualitative changes, which we call transitions.
Work on the Bridge System serves multiple purposes: we have a platform for moving forward toward work on learning and on abstraction; we have developed a set of prototypes for modules reusable by other research groups; we have encountered entirely new, unanticipated problems and opportunities we did not know existed; and we have inspired research focused on related subjects.
The research of the Genesis Group is supported, in part, by the National Science Foundation under Grant Number 0218861.
All of the members of the group believe in some of the following points of dogma; some of the members believe in all of them:
