Russell and Norvig, Chapter 26: Philosophical Foundations

26.1 The Big Questions
- how CAN minds work?
- how DO human minds work?
- can nonhumans have minds?
- weak AI
  - machines can be made to act AS IF they were intelligent
- strong AI
  - machines that act intelligently can have real, conscious minds

26.2 Foundations of Reasoning and Perception
- assumption: brains cause minds
  - also called physicalism, materialism, or biological naturalism
  - opposing view: dualism
- functionalism
  - componential definition of the brain
  - each component defined according to its input/output properties
- mental states (do machines have them)
  - intentional stance
    - mental states inferred from observed behavior
    - non-implementation specific
  - correspondence theory
    - beliefs are structures in the mind
    - beliefs are GROUNDED to sensory evidence
- brain in a vat (BIAV)
  - remove brain from body
  - hook input/output channels to computer
  - computer maintains simulated, fictitous world
  - if theory works for BIAV, then is not necessary for real world

26.3 On the Possibility of Achieving Intelligent Behavior
- can machines think?
  - question is not clearly defined ("think"?)
- Turing
  - can machines pass my intelligence test
  - i.e., convince a conversant that the machine is a human
- mathematical argument
  - computers can't solve halting problems, but humans can
    - halting problem: will a program P halt or run forever
    - humans can "step outside" formalism to answer question
    - computers cannot (?)
  - in general, given a formal system F
    - there exists a sentence G(F) (Godel sentence)
      - G(F) is consistent with F
      - F cannot prove G(F)
    - but humans can prove G(F) by "stepping outside" (?)
    - i.e., some of human reasoning is not algorithmic
  - Godel's Incompleteness Theorem can be derived by a theorem prover
- informality argument
  - human behavior is too complex to be captured by rules
    - but computers are basically rule-based
  - then how do humans do it?
    - humans use technique T
    - so we'll program T
    - but T is complex
    - well, we're pretty good programmers

26.4 Intentionality and Consciousness
- intenionality
  - what beliefs, desires, intentions ARE ABOUT
- consciousness
  - awareness of one's own mental states
- again, an ill-defined property, even of humans
  - Turing proposes that we attribute consciousness to agents that
    act intelligently (unless the agent in a machine)
- chinese room argument [Searle]
  - given a book of rules for Chinese question/response symbol mapping
  - human with no understanding of Chinese can use it to respond
    intelligently, i.e., pass the Turing test
  - neither does the book "understand" Chinese
  - viewing the book as a program implies that even running the right
    program does not generate understanding
  - assuming programs are unconscious
    - can we combine them into a conscious agent ?
    - Searle claims consciousness cannot not be constructed from
      unconscious components
    - but this would rule out brains unless atoms are conscious
  - consciousness depends on the medium, not the program
- brain prosthesis experiment
  - gradually replace brain neurons with electronic counterparts
    - counterparts have identical input/output behavior
    - will consciousness of subject change?

Exercise 26.7
- define "intelligence", "thinking" and "consciousness"