The BCS Machine Intelligence Prize is awarded for a live demonstration of 'Progress Towards Machine Intelligence'.
Congratulations to Emma Byrne and Ken Whelan (University of Wales, Aberystwyth) who won the competition in 2006 and the £1000 Electrolux sponsored prize with their entry Adam, the robot scientist.
MACS, Heriot-Watt University, Ruth Aylett (email@example.com)
3D graphical characters interact in a virtual classroom on a desktop computer: one of the characters is bullied by one of the others in one of a number of possible ways. The victimised character retreats to the library where he asks the user (usually a child aged 10) to suggest what he or she could do to deal with this bullying via free typed input. In the next scene, the advice of the user influences the behaviour of the victim, and the alternation of dramatic scene and advice continues for the time of the demo.
Icogno Ltd, Rollo Carpenter (firstname.lastname@example.org) with the assistance of Televirtual Ltd, Tim Child (email@example.com)
George and Joan are AI characters - literally, as they're quite a handful. They've been learning to be themselves by `borrowing the intelligence' of millions of visitors to the jabberwacky.com website, and from the specialist teaching of their respective mentors. They're quarrelsome, quirky and, sometimes, unexpectedly lifelike. The AI learns more than text: human reactions and emotions. Input becomes output in context, informing the animated expressions and actions of a new breed of avatar - even while you type. Twinning the avatars with real-time speech generation, and lip synch to match, gives added impact. But it's the AI that's in charge.
University of Wales Aberystwyth, Emma Byrne (firstname.lastname@example.org) and Ken Whelan (email@example.com)
Laboratory Automation has made it possible to do experiments in bulk. The amount of data that results from these experiments far outstrips the human capacity to interpret it. Adam addresses this problem. We will show how Adam observing the growth of yeast and using this data, together with knowledge about biology to hypothesise which genes might code for which proteins. Adam then chooses which experiments should be done next, in terms of their informativeness and cost. We will show that Adam can autonomously plan, run and learn from iterations of experiments to derive new knowledge about genetics.
(Image reproduced from The Age, Melbourne, 16/01/04, included with kind permission by Cathy Wilcox)
Cardiff University, Philip Smart (firstname.lastname@example.org), Alia Abdelmoty (email@example.com) and University of Glamorgan, Baher El-Geresy (firstname.lastname@example.org)
Nearly 80% of the information we use is spatially referenced in some way. As humans, we can use our built-in common sense spatial reasoning to relate the information and discover its latent semantics. Oxford is west of London and east of Cardiff; It is between London and Cardiff? SWSRuLE is a system that is able to automatically derive such spatial reasoning rules. It considerably extends our natural spatial common sense to handle large and complex spatial structures. As an example application, the system will demonstrate the deduction of implicit spatial information in large geographic data stores on the semantic web. Also, the system can check the consistency of spatial scenes and is able to trace and report erroneous spatial structures.