Four
REMEMBERING AND LEARNING
Memory as Environment for Thought
p 102
...in the past decade research [...] has been turning more and more to semantically rich domains -domains that have substantial, meaningful content, where skillful performance calls upon large amounts of specialized knowledge retrieved from memory.
p 103
More memory does not necessarily mean more complexity.
SEMANTICALLY RICH DOMAINS
p 103
There is a certain arbitrariness in drawing the boundary between inner and outer environments of artificial systems.
Long-Term Memory
Intuition
How Much Information?
p 108
Even the most talented people require approximately a decade to reach top professional proficiency [mention of Bobby Fisher, Mozart]
Memory for Processes
UNDERSTANDING AND REPRESENTATION
p 111
Before a General Problem Solver can go to work [...], it has to extract from the written statement a description of the problem in terms of constructs that a GPS can deal with [...].
A Program that Understands
p 112
[UNDERSTAND program...] the parsed sentences are examined to discover what objects and sets of objects are being referred to, what properties of objects are mentioned and what are the relations among them, which of the predicates and relations describe *states* and which describe *moves*, and what the goal state is.
Understanding Physics
p 114
ISAAC [a program by Gordon Novak...] has stored in memory information about levers, masses, inclined planes, and the like in the form of simple schemas [...]
A ladder schema, for example, looks something like this:
Ladder Type: ladder Locations: (of foot, top, other points mentioned) Supports: Length: Weight: Attachments: (to other objects)
Size and Simplicity
p 117
We may say that the system becomes more complex because it grows in size, or we may say that it remains simple since its fundamental structure does not change.
LEARNING
Learning with Understanding
Production Systems
p 120
In the past few years a new form of program structure has become popular: the production system. [...] A production system is a set of arbitrarily many productions. Each production is a process that consists of two parts -a set of tests or conditions and a set of actions.
[...] Condition -> Action
Learning from Examples
DISCOVERY PROCESSES
Problem Solving without a Goal
Rediscovering Classic Physics
p 126
[AM and BACON programs] ...discovery processes do not introduce new kinds of complexity into human cognition.