The BCS Machine Intelligence Prize is awarded for a live demonstration of 'Progress Towards Machine Intelligence'.
The competition was held on Tuesday December 13th 2005 at Peterhouse College, Cambridge, UK during the annual SGAI conference AI-2005. Full details of the competition can be found on the Information Page.
Icogno Ltd, Rollo Carpenter (firstname.lastname@example.org)
George is a chatbot personality created with the learning AI found at Jabberwacky.com, and exists only to be an entertaining companion. After a brief exposition of his often-humorous conversational style and range, taken from his many lengthy interactions with visitors online, a member of the audience will be invited to hold an informal chat with him. The outcome will be hard to foretell - he'll use the context of no less than 6.8 million previous interactions to decide what to say. George is a Turing Test contender, and recently won the Loebner Prize as the `most convincingly human' program.
Department of Cybernetics, University of Reading, Will Browne and Dan Scott (email@example.com; firstname.lastname@example.org)
Baron Munchausen's programme intelligently plays Connect 4 in the same manner as humans. People play Connect 4 intelligently without storing libraries of game states or rigid strategies. Patterns are matched so that irrelevant information is discarded, then commonalities in the patterns are abstracted to form higher-level rules. This system is important as it also has this ability - it learns from patterns that it has already learned. These abstracted rules are more compact, efficient and effective and facilitate developmental learning. This is the first example of a learning system that pulls itself up by its boot-straps, which is central to developing intelligent machines.
KDE Group, Queen's University, Belfast, David Bell, Marcel Ono, QingXiang Wu (email@example.com; firstname.lastname@example.org)
A mobile robot (Khepera) demonstrates intelligence as it encounters a new object in its world. It reacts initially in an 'instinctive' way to its first perception of the object - but subsequently it learns that, as this is not a necessary response and as there is evidence to support a more positive one, it can adapt a different interpretation for its purpose, and use it in the future. The acquired knowledge combined with modest innate knowledge is used for uncertain reasoning. Inflexible standard evidential reasoning methods are enhanced by qualitative assessment of evidence and adjustment of knowledge/behaviour.
BT, Simon Thompson (email@example.com)
Web service composition and intelligent portal building system. Kreno uses AI planning, goal semantics, scheduling and an autonomous agent to compose web services together on demand to provide users with resolutions for their requests. It is designed to be easily integrated with web technology and is supported by a set of analysis and debugging tools.