Announcing an ICML Workshop on

Predictive Representations of World Knowledge

Proceedings

Rationale

It has long been postulated that knowledge of the world ultimately comes down to predictions about what we will sense as a function of what we do. For example, in this view we know what is inside a box if we can predict what we would see if we were to open it. We know the three-dimensional shape of an object if we can predict its new appearance for each of various ways in which we might rotate it. We might even be said to know that "Tweety is a bird" if we can make appropriate predictions about what we might see and hear (e.g., about feathers, flying, and chirps) if we were to meet Tweety. Such predictive representations of world knowledge have important potential advantages. If predictions are defined in terms of primitive sensations and actions (i.e., are grounded), then they can be directly compared with what actually does happen, enabling the knowledge to be verified or perhaps even learned without human intervention. If the predictions are deterministic, or Markov in an appropriate sense, then the knowledge can be immediately used in a variety of state-space planning methods.

For these and other reasons, understanding world knowledge in predictive, sensori-motor terms has been a long-standing goal of philosophy, psychology, and artificial intelligence. So far it has remained a distant goal, but recent progress in machine learning seems to bring nearer the possibility of addressing it productively with mathematics and computational studies. In particular, we are thinking of research on:

The goal of this workshop is to bring together scientists interested in these and other topics to present, discuss, and make further progress toward understanding how knowledge of the world can be represented in predictive, sensori-motor terms.

The format will be a mix of presentations and discussion. Each speaker will be given a smallish amount of time, perhaps 20 minutes depending on the number of presentations. About half of the total workshop time will be reserved for discussion, which may be questions for a speaker, discussion of topics related to one of the talks, or introduction of new topics. The allocation will be done in realtime at the discretion of the organizers.

Program Committee

Participation and Submissions

Participation will be by invitation from the program committee based on a small written submission. Think of the submission as either 1) an abstract for a talk you would like to give at the workshop (though of course there will be time for only a few), 2) a statement of interest and ideas relevant to the workshop's goals, or 3) a short paper or extended abstract on your relevant work (this may be published or about to be published). The submissions will help us to keep the workshop focused, reasonably small, and yet enable the workshop to be open to new work. Final versions of the accepted submissions will also be published in an online workshop proceedings. Submissions must be received no later than March 26, 2004. Email notification of workshop attendance decisions will be sent out by April 16, 2004.

Please send submissions to both Rich Sutton (sutton at cs dot ualberta dot ca) and Satinder Singh (baveja at umich dot edu) for distribution to the committee. All submissions must be made via email with PDF format preferred.

Important Dates

March 26, 2004 Workshop submission deadline
Apr 16, 2004 Notification of participants
May 7, 2004 Deadline for inclusion in the workshop proceedings
July 8, 2004 Workshop date

Proceedings

Link to related workshop on Anticipatory Behavior in Adaptive Learning Systems

Additional ICML information here.

This is prwk.rlai.net.