White paper and slides prepared for
DARPA Workshop on Cognitive Systems

Airlie House, Warrenton, Virginia, USA, 3-6 Nov 2002.
by Aaron Sloman
Based partly on or in reaction to discussions at the workshop.

Note added 23 Aug 2007

I have found that the web site used for this workshop and all the online papers are no longer available. I have saved copies of several of them, however.
Unfortunately the links below to DARPA sites no longer work.

For information on the DARPA Cognitive Systems initiative see Ron Brachman's paper 'Systems that know what they are doing'
http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1134363
http://doi.ieeecomputersociety.org/10.1109/MIS.2002.1134363

in Intelligent Systems Volume: 17, Issue: 6, Nov/Dec 2002, pages 67- 71, ISSN: 1541-1672
A plain text version freely available.
A related presentation

1. White paper on architectures, submitted in advance of the workshop:
How To Think About Cognitive Systems: Requirements and Designs

Three formats available:

2. Modified version of my (partly critical) slides prepared for panel on Monday 4th Nov, on
The architecture of an enduring, perceiving, learning, cognitive agent:
How to think about architectures: Requirements, designs, niche space, design space, tradeoffs, trajectories architecture-based conceptions of emotions and other affective states.

All the white papers submitted to the workshop were available at these locations, but the site appears to have been disabled.
http://www.dsic-web.net/meetings/oy8guwod/papers.html

Slide presentations from the workshop were originally here (now not available):
http://www.dsic-web.net/meetings/oy8guwod/presentations.html

NOTE added 23 Aug 2007

The people at the workshop rejected my proposals in favour of what they thought DARPA was more likely to fund, namely a project to produce an intelligent personal assistant. Such a project is defined here:
http://www.darpa.mil/ipto/programs/pal/index.htm
Personalized Assistant that Learns (PAL)
"The mission of the PAL program is to radically improve the way computers support humans by enabling systems that are cognitive, i.e., computer systems that can reason, learn from experience, be told what to do, explain what they are doing, reflect on their experience, and respond robustly to surprise. More specifically, PAL will develop a series of prototype cognitive systems that can act as an assistant for commanders and staff. Successful completion of this program will usher in a new era of computational support for a broad range of human activity."
Movie illustrating the PAL concept
I argued that that sort of goal cannot be achieved without first learning how to build child-like robots that understand the basics of being an agent in a 3-D physical world, and then go on to learn many other things on that basis, as humans do. Trying to go straight for an 'adult' expert which lacks that foundation will produce something very brittle and hard to extend.

UK Computing Grand Challenges Initiative

A UK workshop to discuss computing grand challenge proposals was held in Edinburgh Nov 22-24th 2002. One of the grand challenge proposals emerging from the workshop and subsequent related activities is the 'Architecture of Brain and Mind' proposal. Information about that proposal and links to other grand challenge proposals can be found here: http://www.cs.bham.ac.uk/research/cogaff/gc

This includes some comments on possible metrics for progress on the kind of project being considered.

White paper and slides prepared for DARPA Workshop on Cognitive Systems

Airlie House, Warrenton, Virginia, USA, 3-6 Nov 2002. by Aaron Sloman Based partly on or in reaction to discussions at the workshop.