COGNITIVE SCIENCE

FINAL TAKE-HOME EXAMINATION

DUE: Monday, November 7. 2005. 5pm.

WORD LENGTH: 2100 (TOTAL)

Answer ONE question from EACH of the three sections. Bibliographical references or page numbers to the articles in the course reader are required but it is not necessary to consult texts other than those provided in your reader. Make sure you explain the argument you are considering carefully, use examples to demonstrate your understanding, and provide reasons as to why the argument should be accepted or rejected.

Attach a completed cover sheet to your exam and deposit in the slot outside the main office in E 306, Forgan Smith. The coversheet can be downloaded from http://www.uq.edu.au/hprc/documents/cover.pdf

SECTION 1. 

1.     It is often argued that for artificial minds to be possible, the functionalist theory of mind must be the correct one. Explain the connection between functionalism and the possibility of artificial intelligence. It is also argued that the possible multiple realizability of mental functions is the best reason for endorsing functionalism. What does it mean to say that the mind is multiply realizable and is it a good reason for endorsing functionalism?

2.     What does it mean to say that the functions the mind can perform are computable functions? Is it likely to be true that all the functions the mind performs or is capable of performing are computable functions?

3.     What is Searle’s ‘Chinese Room’ thought experiment supposed to show us about artificial intelligence? Does the thought experiment support his conclusions?

SECTION 2.

1.     What are the basic principles of neural networks and how do they differ from classical machines? It is often claimed that such machines are more likely than classical machines to warrant being described as intelligent because they are capable of ‘learning’. Evaluate this claim.

2.     Jerry Fodor has argued that because language is systematic, thought must be also, and since connectionist machines are not systematic, they cannot model thought as well as classical serial machines. Evaluate the argument against connectionism from systematicity.

3.     What is the epistemological ‘frame problem’ in artificial intelligence? What obstacles, if any, are there to a formal or information theoretic solution (one given in terms of rules for deriving relevant conclusions from available information) to the problem?

SECTION 3.

1.     What is it that explains the fact that a mental state is conscious according to Higher-order Representational Theories) of consciousness (such as HOP and HOT)? What is the relevance of this approach to consciousness for research in AI? Does the theory succeed in defining what makes a mental state conscious?

2.     Chalmers argues that it is only the ‘easy’ problems of consciousness that are dealt with in neuroscience and cognitive science, and that the ‘hard’ problem requires revising our understanding of the fundamental properties of the natural world. Is consciousness a ‘hard’ problem in the way Chalmers suggests?

3.     Jackson, among others, has argued that once one takes into account all the physical and functional facts about experience, there is something left out. Explain and evaluate his ‘knowledge argument’ for this conclusion. If he is right, would it be impossible for there to be an artificial subject of conscious experience?