The BCS Machine Intelligence Prize is awarded for a live demonstration of 'Progress Towards Machine Intelligence'.

Summary of Finalists - 2012

Photographs from the competition     Pre-competition photographs

RAGA: Rapidly Adapting Game Agent

Simon Lucas (sml@essex.ac.uk) and Diego Perez (dperez@essex.ac.uk), University of Essex

Designing software agents that can learn to play games to a high standard without being explicitly programmed to has been a grand challenge since the dawn of AI research. Recent research combining selective tree search with reinforcement learning can be used to enable the automatic construction of competitive game agents. This allows designers or end users to produce novel strategy games, e.g. for mobile apps, and immediately play them against a learning agent without having to program it. The demonstration will show construction of a novel game and the agent rapidly learning to play it.


George and Mary

Faisal L. Kadri (faisal@artificialpsychology.com , http://artificialpsychology.com), Montreal, Canada

George and Mary is a dialogue player made of two artificial personalities. The personalities exchange sentences selected from a repertoire organized in four dimensions and increment motivational state with each exchange. The dialogue of Hansel and Gretel is used to demonstrate how the fable can be made to appeal to all ages. The demonstration will show how age preferred sentences are selected by following fuzzy logic rules, and show the creation of threads of conversation and compatibility with other personality models. Finally, a demonstration of work in progress which will lead to the formation of a single personality from combining the two; Mary acting as the Ego and George as the alter ego or artificial imagination, where the artificial personalities of all friends and foes reside and referenced as behavioural norms.


Debugging and simulating tool for tactical resource
planning

Ahmed Mohamed (amhmoh@essex.ac.uk), University of Essex/ British Telecom

This system is used to animate the plan done by a planning algorithm. It has the ability to change in the environment by adding, editing or removing resources and tasks and see how the system will handle these changes. This system can also be used as a debugger to test different scenarios that can be created manually by changing resources and tasks position or randomly by letting the tool create the problem. The plans can be loaded from a database as well.


Storytelling chatbot - HMS Ark Royal

Collette Curry (collette.curry@stu.mmu.ac.uk), Manchester Metropolitan University

Featuring ships named HMS Ark Royal dating from 1588 to 2011 and introduced by ‘The Narrator’- a talking AI entity that learns from conversation, prompting discussion, interaction and offering stories from the eras, ships and the people that served on them. Creative writing produces text which is fed into the corpus, becoming part of the story. This combined with each conversation makes stories unique to each participant and encourages use of the system. For each user, the system starts up where they last used it and continues the story where it was ended. Further stories are installed locally for use offline.


Neural Cognitive Robot

Tang Huajin (htang@i2r.a-star.edu.sg), Tian Bo, Tan Chin Hiong, Li Haizhou, Institute for Infocomm Research, A*STAR, Singapore; Tan Kay Chen, Department of Electrical and Computer Engineering, National University of Singapore

The robot is fully autonomously moving around the unknown maze environment and simultaneously mapping the maze into a path map. The robot is able to form an internal representation of the external environment by cognitive memory mechanisms. To demonstrate the effectiveness of the neuro-cognitive robot, we can change the environment, e.g. the walls and the object locations in runtime, and show that the robot can perform the tasks in a new environment. Imagine what a rat can do inside a maze, our Neco can do better! Such capability of the robot enables many new applications in future.