Machine consciousness

In the last few years a new discipline has begun to emerge: machine 
consciousness. This talk will describe the background to this movement, 
and will present a line of thought showing how the problem of constructing 
a truly autonomous robot may also constitute an approach to building a 
conscious machine. The basis of the theory is that an intelligent robot 
will need to simulate both itself and its environment in order to make 
good decisions about actions, and that the nature and operation of the 
internal self model may well support some consciousness-related phenomena. 

As part of an investigation into machine consciousness, We are currently 
developing a robot that we hope will one day possess and use a self-model 
similar to our own. We believe that this requires a robot that does not 
merely fit within a human envelope, but one that is anthropomimetic - 
with a skeleton, muscles, tendons, eyeballs, etc. - a robot that will 
have to control itself using motor programs qualitatively similar to 
those of humans. The early indications from our prototypes are that 
such robots are very different from conventional humanoids; the many 
degrees of freedom and the presence of active and passive elasticity 
do provide strikingly lifelike movement, but the control problems may 
not be tractable using conventional robotic methods. We have also 
developed a framework for the internal models of self and environment - 
a set of complex physics-based simulations capable of representing the 
robot's interactions with the environment with whatever degree of precision 
is required. The talk will describe the current status of the work, and 
will explore strategies for integrating the models into the physical 
robotic system.