AI23 - Bioinspired Computing
AI23 - Intelligence Without Reason - Reading Guide

...back to the module home page.
While reading Brooks' (1991) "Intelligence without reason", remember, it is:

1. Introduction

  • Intelligence is defined loosely by Brooks as "the stuff that humans do, pretty much all the time". Do you think this is a fair description?

  • What is the difference between the "top down" and "bottom up" approached to AI that Brooks refers to?

    2. Robots

    [It may be easier to understand this section once the whole paper has been read through once.]

  • What is so important about "situatedness" and "embodiment"? What do these terms mean?

    3. Computers

    3.1

  • Why did Turing think chess was an interesting problem?

  • Why is Brooks unimpressed by the excellent performance of today's artificially intelligent chess players?

  • What is the Turing test, and what, from Brooks perspective, is wrong with it?

    3.2

    What does Brooks think is wrong with the statement "intelligence is mostly heuristic search"? Why might AI researchers have been led to believe this statement?

    3.3

    [Do not worry if this section is hard to follow]

    3.4

  • How was AI's descent into "abstraction" justified? Was this justification legitimate?

  • Is "connecting to the real world" a problem for engineers, rather than AI researchers?

    3.5

  • How does "embodiment ground knowledge"? (This is a very hard problem, don't worry if you don't understand the question).

    3.6

  • Why did "the need to get things operational" cause a "shift of approach" by the Autonomous Land Vehicle project?

    • 3.7

  • What is the main argument in this section? Does this argument generalise from vision to a variety of sensory inputs? If so, does it agree with the example of Barbara Webb's study of cricket behaviour?

    3.8

    [Return to this section when we discuss artificial neural networks. (In particular, Lecture 4 will introduce this material)]

    4. Biology

    [For our current purposes, this section is not as important as others]

    5. Ideas

    5.1

  • When and why is "the world its own best model"?

    5.2

  • How does the embodiment ensure that the world grounds regress?

    5.3

  • Brooks claims that since evolution had spent billions of years sorting out basic behavioural competences (such as moving around without bumping into things) before the relatively recent arrival of 'higher-level' intellect, AI researchers should study these basic competences. Are you convinced?

    5.4

  • Is Brooks clear about what he means by "emergence"?

    6. Thought

  • What is the motivation behind Brooks' "principles"?

    6.2

  • What is it to be a reactive system? When is being reactive beneficial? When is it a hindrance?

    6.3

  • What are examples of "deictic representations"? How do they differ from non-deictic representatios? Which type are more often found in traditional artificial intelligence systems?

    6.4

  • Will Brooks' approach scale? What does this question mean and how important is it?

    7. Conclusion

  • What does Brooks mean by "Intelligence without reason"?

    ...to the top of the page.
    ...back to the module home page.
    Netta Cohen
    Email: netta@comp.leeds.ac.uk