Background
Symposium at AISB'06
Grand Challenge 5: Architecture of Brain and Mind

This symposium has a number of sources, but arises mainly out of an initiative of the UK Computing Research Committee (UKCRC), begun in 2002, promoting identification of research Grand Challenges. Information about the UK Grand Challenges initiative is available here, including the list of current proposals and an invitation to submit new grand challenge proposals. A booklet edited by Tony Hoare and Robin Milner presenting the current collection of Grand Challenges is available online (PDF).

One of the challenges is GC5: Architecture of Brain and Mind. See also http://www.cse.dmu.ac.uk/~aayesh/gc5/.

The main thrust of GC5 is to promote a combination of bottom-up, top-down and middle-out research into the nature of minds (human and animal) and the mechanisms in brains, and the ways in which brains can support the existence and functioning of minds. From this viewpoint minds are virtual machines running on brains, which are physical machines. The most important feature of minds is that unlike most other machines which merely manipulate matter and energy, minds manipulate information -- which they acquire, store, analyse, transform, generate, communicate, interpret and use in many ways.

For further discussion of the notion of information processing in virtual machines see this presentation (PDF).

There are many ways in which research on this challenge can be motivated and progress measured. One way that has been selected as particularly important aims at design and implementation, fifteen to twenty years from now, of a complete functioning robot with a significant (even if very small) subset of the capabilities of a young child, somewhere between one and five years old.

The challenge is to integrate diverse capabilities, including visual and other kinds of perception, motor control, intelligent physical manipulation, problem-solving, planning and executing various kinds of actions, some social interaction, some linguistic competence, some subset of motivational and other affective states, and some self-understanding, along with the ability to learn and develop in ways that extend these capabilities.

The Grand Challenge project involves working in parallel on

  1. understanding and modelling brain mechanisms that are capable of supporting such diverse processes, and
  2. designing, implementing and testing (initially using whatever computational tools are available) high level models for the capabilities to be replicated in the robot
This requires a mixture of bottom-up and top-top down research, with researchers frequently talking to each other about the issues.

The hope is that many aspects of such a robot would be closely related to explanations of how humans (and some other animals) work, and how they learn and develop. Such a robot cannot be achieved in the near future, but the long term goal will help to inspire and constrain short-term and medium-term research and provide a basis for assessing progress, especially if researchers work backwards from requirements for the long term goals, instead of only extending what they already know, as most researchers do.

No specific funding has been allocate to support GC5, though UK Research councils have indicated that they will look favourably on applications inspired by the grand challenge proposals, and some have already been funded. (Some of the ongoing work will be presented at the symposium.) There is also a large amount of EC funding in the Cognitive Systems initiative for work in this area. A first wave of projects started in 2004, including the CoSy project whose members are helping to organise this event, and a new set of projects has recently been selected (which should be announced at the Cognitive Systems Projects page), one of which is the euCognition network project funding this symposium

Some suggestions for encouraging cooperation between researchers with different approaches, and for identifying milestones against which progress can be assessed were made in this document (AI in a New Millenium: Obstacles & Opportunities).

GC5 is one of two biologically inspired computing challenges in the UKCRC list, the other (GC1) being an attempt to to build a working model of all the neural circuitry of a simple animal as described here. Of the other Grand Challenge proposals the one most closely related to GC5 is G7: Journeys in Non-Classical Computation since it is likely that new forms of computation will eventually needed for Grand Challenge 5, and research into minds and brains and how they are related will be a major driver for research on new forms of computation (broadly construed as processing of information, whether done by what we now call computers or by other mechanisms).

This symposium is organised by members of the GC5 Committee, with the support of the euCognition Network and Birmingham members of the CoSy Project, both funded by the EC FP6 Cognitive Systems Initiative,

An earlier event organised with similar motivation was the Edinburgh Tutorial on Representation and Learning in Robots and Animals, at IJCAI'05. The booklet of presentations and papers is here. Those who attended found this a very exciting and stimulating event because of its focus on integrating research activities normally pursued separately. One of the themes was that solutions that work for a narrow problem often cannot easily be integrated into a system with broader scope. The same issue will arise in this symposium.


This site is maintained by Aaron Sloman
http://www.cs.bham.ac.uk/~axs/
Last updated: 4 Mar 2006