Action is a process of acting or doing.
Examples of an Action:
Scheduler and making
decision routine define which action should be executed next.
The most
probable candidate for the action is concept with the highest
desirability value.
Action is a concept
which affects a hardcoded unit’s behavior.
Example:
“Run Programmator routine” is action which affects hardcoded unit behavior.
(The hardcoded unit affects the World
state. Changes in the World state lead to events).
Example:
“Take
lunch” action causes:
Thought results in action.
Example:
“I need good history knowledge” thought results in “Read a history book” action.
Strong AI considers every action as part of an experiment.
Event Correlation Analyzer searches
for cause-effect relations between
action and subsequent event.
Example:
“Read a history book” – is an action.
“Read a history book” is a cause concept for “good history knowledge”
concept.
“Good history knowledge” is an effect
concept for “Read a
history book” action.
Event Correlation Analyzer reveals correlation between “Read a history book” and “good
history knowledge’” and creates cause-effect
relations “’Read a history book’->’ good history knowledge’”.
Every action
is evaluated by the reward
distribution routine. As a result of such evaluation concepts
responsible for the action are updated. In particular, desirability attributes of these concepts are updated (by reward
value).
Example:
A man
says “I love you” to a girl. The girl kisses the man. The man enjoys the kiss.
That means that the man got a reward for “Saying ‘I
love you’” action. Reward of the kiss is
calculated accordaning to the man’s super goals (desire
for tenderness, sexual instinct). The concepts / sub goals responsible for “Saying ‘I love you’” action
is rewarded. The desirability attribute
value of the concept is increased. “Saying ‘I love you’” action / sub goal becomes more desirable.