CONTENTS
NEWS
24 Jan 2004: Additions from Darryl Davis for
the links file.
23 Jan 2004: Added post-workshop
summary by Aladdin Ayesh
31 Dec 2003: programme revised
30 Dec 2003: Sections added on car parking and abstracts below.
PURPOSE OF WORKSHOP
This workshop is meant to be preparation for a larger conference
organised by Tony Hoare and Robin Milner on grand challenges to be held
in Newcastle, March 29-31 2004, co-located with the CPHC annual
conference. Details and news will be posted here:
http://www.nesc.ac.uk/esi/events/Grand_Challenges/
http://www.nesc.ac.uk/esi/events/Grand_Challenges/gcconf04
The March conference will be concerned with all the proposed grand
challenges. Each of them will have some time allocated for discussion.
Aaron Sloman has been asked to give a presentation on GC5
Architecture of Brain and Mind. The purpose of the workshop on
January 5th is to help publicise this particular grand challenge,
recruit potential participants, and collect ideas to be presented at the
March conference and in documents on the main Grand Challenges website,
above.
2000-word submissions to that conference are invited from interested
researchers. Deadline for submissions is 1st February. Details are
available at the Grand_Challenges web site. The January workshop should
help to provide useful background information for anyone planning to
submit a paper relating to Cognitive Systems or Cognitive Science. It is
expect that most papers will be concerned with developing one of the
existing grand challenge proposals, though new ideas that meet the
criteria for a grand challenge may also be accepted.
BACKGROUND TO WORKSHOP: Grand Challenge 5
There is an overview of what the 'Architecture of Brain and Mind' grand
challenge is about here,
http://www.cs.bham.ac.uk/research/cogaff/gc/
including pointers to the main grand challenge website at NESC and
various other relevant items. Anyone can read the archived email
discussions of GC 5 at the NESC site:
http://archives.nesc.ac.uk/gcproposal-5/
Anyone can join the mailing list. You can't post to it unless you join,
though people already on the list can forward comments and suggestions.
Letter from Robin Milner and Tony Hoare to GC Moderators, October 2003
ORIGINAL SUBMISSIONS TO GRAND CHALLENGE WORKSHOP NOVEMBER 2002
All the original submissions to the first 'Grand Challenge' workshop
held in November 2002 in Edinburgh are avalable online. The submissions
assigned to 'Panel D' at the workshop, from which the 'Architecture of
Brain and Mind' proposal emerged, are available here:
http://www.nesc.ac.uk/esi/events/Grand_Challenges/paneld/index.html
The list of papers gives titles of the documents but not authors. Some
of the documents specify authors, but not all do.
The other submissions were handled by other panels,
WORKSHOP ORGANISERS
The January workshop is being organised by
With local help also gratefully received from
Bob John (rij
AT dmu.ac.uk)
If you would like to attend, please email both of the organisers
as soon as possible. If you have suggestions for the agenda please
include them. The workshop can accommodate up to about 40 participants
on a first-come first-served basis. There are no funds for travel, and
there may be a small registration fee -- to cover coffee, cost of
facilities for the workshop, etc. Further details will be sent to
accepted participants.
LOCATION MAPS AND PARKING
Centre for Computational Intelligence,
School of Computing at De Montfort University
The Gateway, Leicester LE1 9BH
Campus map
The meeting is in the Gateway building, labelled 13 on the map.
Directions for getting to the campus
PARKING (Added 30 Dec 2003)
Aladdin has managed to book a small number of parking places at
the university car park. Please let him know
immediately if you need one. (First come first served).
There is also a pay car park opposite the gateway house building
(where our meeting is). If that is full, the next nearby car park
is the phoenix car park near Welford Road (the car park entrance
may be on New Walk Street). These roads should appear on the University
maps.
Additional car parks are available around the city centre near most of
the hotels mentioned below (Holiday Inn, St James, and Ibis Hotel).
LUNCH
During the 90 minute lunch break, participants can buy food in the
brand new Campus Centre which includes a food hall (Building 3 on the
campus map)
ACCOMMODATION
For people who need accommodation, the following information has been
provided by Aladdin Ayesh and Michelle Dale at DMU:
There are several hotels and guest houses around the University and the
train station (mostly walking distance to both). These are few examples:
-
Holiday Inn Hotel 0116 251 3169 (City Centre, close to
University)
-
Saint James Hotel 0116 251 0666 (City Centre)
-
Ramada Jarvis Hotel 0116 257 5500 (Near train station, en route
to University)
-
Ibis Hotel 0116 248 7200 (Near train station)
-
Abinger Guest House 0116 255 4674
-
Westcote Lodge 0116 254 5811
-
Bradgate Hotel 0116 2557993
near train station.
The usual or minimum cost for guest houses is £30 per night, for hotels
£50 per night.
ABSTRACTS (Added 30 Dec 2003)
Short abstracts for the presentations are now
available here.
If anyone attending the workshop or anyone on the GC5 mailing list,
wishes to send
me
a paper in pdf, html, or plain text (no word or rtf or
powerpoint files please) (preferably no more than 20 pages, and
preferably no more than 1Mbyte in size, I'll make it available at this
site, provided that you also give me a short abstract of at most 100
words (plain text or html only). The paper, or at least the abstract,
should give your name, affiliation and an email address or url.
(I don't want to have to answer questions about how to get in touch with
people.) The paper must include your name (unlike several papers
submitted to the original Grand Challenge workshop).
If you prefer, just send me the abstract with a url to the paper if it
is already online.
All such submissions should address the main theme of the Grand
Challenge, namely how to put together all the pieces studied separately
in different subfields of AI, of Psychology, of Neuroscience, etc. i.e.
requirements for combining all the different capabilities of a typical
human (or intelligent animal).
[Note: PDF produced by latex should use the times font (\usepackage
{times}) rather than the default latex font (CMR), which produces PDF
files which look blurred and difficult to read in acroread.]
DRAFT AGENDA FOR THE WORKSHOP:
Last modified 30 Dec 2003.
May change again in the light of further suggestions
and comments received.
9.30-10.00: Arrival/Coffee/Tea etc.
10.00-10.05 Welcome and information about arrangements for the day.
10.05-10.30 Introductions
People all say briefly who they are and what their interests are.
(Assuming that we will not have more than about 20 people
present. If the number is much larger, this may be infeasible.)
10.30-11.00
A.Sloman (with interruptions and help as needed):
Overview of the UK Grand Challenge initiative (5-10 mins)
and the history and status of Grand Challenge 5 (Architecture
of Brain and Mind) (about 20 mins), including summary of
controversies and points of agreement in the email
discussions so far:
http://archives.nesc.ac.uk/gcproposal-5/
11.00-11.30
Everyone: relevant initiatives elsewhere.
The point of this session is to establish the national and
international context and especially scope for international
collaboration.
It may be useful for people who were involved in 'grand
challenge' proposals submitted to the EU FP6 last month
to summarise what they were proposing to do.
11.30-12.30
Presentations and discussion
For more details on the presentations listed here see
the abstracts.
Joanna Bryson (Bath):
Short presentation on relations between AI
architectures and neuroscience.
Edmund Furse (imitation.uk.com):
Short presentation on research methodology, with particular reference to
Cognitive Science and learning.
Murray Shanahan (Imperial College)
Intelligence and Imagination: Why We Need Both for Grand
Challenge 5
Mark Lee (Aberystwyth)
Developmental Learning and its importance for Brain and Mind
William Edmondson (Birmingham)
Building an Intelligent System with a Brain.
In remaining time:
Discussion of objectives, themes, possible time-tables
controversies about what should and should not be in the
challenge (see the GC5 mailing list discussion).
12.30-2pm Lunch and informal discussion
2.00-2.15 pm
Sophie Kain (Thales):
A proposal for networking related to the Grand Challenge.
Possible final sessions
The following agenda items may be substantially revised in the light of
the morning discussion.
2.15-3.00
Everyone:
Brainstorming attempt to collect a first draft set of major
headings and sub-headings to go into a grand challenge proposal,
to succeed these documents:
3.00-3.45 pm
Begin discussion of list of major gaps in current knowledge relevant to
the Grand Challenge.
Attempt to identify major gaps or unsolved problems
in our current knowledge and understanding.
Example headings for list of gaps:
- What are the main functions of the various forms of perception?
E.g. what are affordances, how are they pereived, represented and used?
How do different sensory modalities (touch, smell, hearing,
proprioception) work together.
What are the requirements for perception of causation, of functional
relations, of intelligent action?
- Varieties of actions:
How they are initiated, executed, controlled, modulated, terminated,
combined with other actions, represented, reasoned about, ...
What dimensions of variation are there, e.g. innate vs learnt, automatic
vs deliberative, familiar vs novel, continuous vs discrete, ballistic vs
controlled by feedback, different levels of abstraction, different
time-scales, communicative vs non-communicative,
- Natural language:
What are the interactions between language, perception and action?
How many different sorts of use of language are there, e.g.
communication, vs thinking and reasoning.
[For some of us, language is mostly used for thinking,
planning, etc!]
- Varieties of learning and development
How many varieties of learning and development are there?
Learning to recognize and classify, to react, to reason, to communicate,
to control actions, to monitor actions, to monitor thought processes, to
think about novel categories (ontological extension), to have new goals,
preferences, values, to make ethical and aesthetic judgements, etc.
What forms of learning and development occur in different parts of the
architecture?
What innate forms of representation, capabilities, ontologies, knowledge
are required to enable human-like learning and development in a
typical human environment?
- Varieties of self-knowledge/self-understanding/self-control
e.g. in different sorts of robots, different sorts of animals,
other kinds of machines.
What are the implications of self-knowledge in an
information-processing system (e.g. what are
requirements for thinking about about varieties of
information processing, in oneself and others)?
- Varieties of understanding of space, spatial structure, motion,
action, causation?
How do these kinds of understanding relate to what one can do
in space, on various spatial and temporal scales?
How many different kinds of causation does a human-like system
need to understand?
(E.g. contrast requirements for hunting, grazing, tree-climbing,
flying, swimming, nest-building, animals.)
- Motivation, emotion, moods, attitudes, and other varieties of
affective states and processes
- Varieties of consciousness
of things in the environment, of one's goals, of what one is
doing, of what one knows, of what one is perceiving, etc.
- Varieties of knowledge required by human-like systems
how this develops and changes from infancy to adulthood
what ontologies are used in different stages of development
what forms of representation and reasoning are used at different
stages
- Knowledge about how to reason
e.g. the development of
mathematical, logical, and other kinds of 'topic neutral'
knowledge.
- Kinds of social knowledge and skills
- Varieties of forms of representation
How many varieties does a human-like system need:
neural, logical, linguistic, visual/spatial,
procedural, rule-based,
dynamical-systems,
etc....
How many different types of mechanisms are required for
processing information of different forms, for different
purposes, in different parts of a
complex architecture,
- Varieties of creativity
- Architectures and requirements for architectures
Do we understand the variety of possible architectures for
intelligent systems (natural and artificial)?
('Architecture' = the functional decomposition of a
complex system with many interacting components)
Can we produce a good language/ontology for specifying
architectures and the requirements against which they should
be evaluated?
- What is not known about typical human information processing
architectures?
(There's much speculation in psychology,
psychiatry, neuroscience, philosophy ....)
What don't we know about what biological mechanisms
(e.g. neural mechanisms) can actually do?
- As designers or scientists can we specify and reason about a
complex architecture with many components closely coupled to a
complex ill-understood dynamic environment including intelligent
agents?
- What don't we know about how to build physical components for
human-like or animal-like robots, and how the physical
properties determine information-processing requirements
for using and controlling body parts.
- Do we know what tools are required to support research on the
grand challenge?
Do appropriate tools exist, or will we need new kinds?
What are requirements for such tools?
(E.g. good support for rapid
prototyping, extendability, open code, portability, support for
both traditional symbolic AI and other types of
representations and mechanisms, support for concurrent subsystems
with rich interactions...)
A related draft list of knowledge gaps can be found here:
http://www.cs.bham.ac.uk/research/cogaff/gc/ai-cogsys-gaps.html
Suggestion:
A 'Grand' challenge project needs to work backwards from a
distant goal about which we have many gaps in knowledge, instead of
taking what we already know how to do and proposing extensions and
applications. I.e. 'backward chaining' not 'forward chaining' should
dominate project planning, though both are required in the detailed
work.
3.45-4.00
Break, if needed.
4.00-4.30 pm
Discussion of next steps, including, possibly:
- How to recruit more participants in GC5 discussions (very
few people have so far subscribed to the gc5 email list:
only 17 when I last checked).
- Plans for the CPHC/UKCRC conference in March 29-31 Newcastle
(A.S. has been invited to present GC5, but it should not simply
be a personal statement: there must be a research community
behind it.)
- Whether there should be more meetings, more collaborative
attempts to refine the research goals, etc.
- How to produce more detailed overviews of
(a) gaps in current knowledge
(b) possible intermediate goals to aim for
(c) who is already doing relevant work in the UK and
what they are doing
- Whether to link up with AISB, ECCAI or some other such
organisation(s) that are, or should be, interested in this
sort of grand challenge. Are BCS and UKCRC enough?
What about other disciplines, e.g. British Psychological
Society, Experimental psychology society, etc.?
etc.
LINKS TO RELATED INFORMATION
URLs contributed by interested participants are available
here
Last updated: 3 Jan 2004
Aaron Sloman