( The Development and Analysis of Conscious Machines )

Thesis

The complete thesis is available here: Gamez08_MC-Thesis.pdf (5 MB). You will need to install Adobe Reader to view it. To download the document, right click on the link and choose "Save Target As" or "Save Link As". Individual chapters can be viewed by clicking on the links in the contents.

Abstract

This PhD was carried out as part of the CRONOS project and one of its main achievements was the development of a method for predicting and describing the conscious states of artificial systems. This could help machine consciousness to become more scientific and it could also be used to make predictions about the consciousness of biological systems.

To demonstrate this methodology, a spiking neural network was developed to control the eye movements of the SIMNOS virtual robot. This network learns the association between sensory input and motor output and uses this knowledge to 'imagine' the consequences of different eye movements and avoid stimuli that negatively affect its 'emotions'. This network exhibits a limited form of conscious behaviour, has some of the cognitive characteristics associated with consciousness, and Tononi's, Aleksander's and Metzinger's theories of consciousness were used to make detailed predictions about its phenomenal states.

The spiking neural network was modelled using the SpikeStream simulator, which was developed as part of this PhD and can simulate up to 100,000 neurons. SpikeStream has good performance, a comprehensive graphical interface and it can send and receive spikes to and from real and virtual robots across a network.

This thesis makes a number of theoretical contributions to the study of natural and artificial consciousness, which include a discussion of the relationship between the phenomenal and the physical, a distinction between type I and type II potential correlates of consciousness, and an analysis of conscious will and conscious control. The different areas of machine consciousness research are also classified and some of the challenges facing work in this area are covered in detail.

Contents

Click on the "+" symbol next to each heading to view summaries of the chapters, which can be downloaded by clicking on the appropriate PDF link.