Behaviors as the building block of intelligence

Behaviors are a mapping between sensory data to some motor actions in order to achieve a task. These can be divided into three types:

  • Reflexive behaviors: stimulus-response, innate. These can be further divided into:
    • Reflexes: the response last as long as the stimulus does.
    • Taxes: the response is to move to a certain orientation.
    • Fixed action pattern: the response continues longer than the stimulus.
  • Reactive behaviors: learned and then consolidated through repetition to become unconscious.
  • Conscious behaviors: deliberative behaviors

Behaviors are coordinated and controlled by an innate releasing mechanism, which presupposes a specific stimulus for a certain response.

We can use schema theory 20211229200605 as a tool for mapping sensing and action on behaviors.

Behaviors study served as a key to introduce a new paradigm on the development of robotics 20211229192654.

Perception

On behaviors, perception serves two major functions: to release behaviors and to provide information needed for a behavior to be accomplished. Neisser defined two different types of perception systems on the brain:

  • Direct perception: uses affordances (perceivable potentiality on the environment), low structures on the brain. For example, detecting red color.
  • Recognition: need mental models to distinguish information. Recognize some chair as my chair.

How behaviors are learned

  • Innate: born with it.
  • Sequence of innate: also worn with them, but they are sequenced.
  • Innate with memory: born with them, but they need some kink of initialization, such as the environment.
  • Learned: they are not born with them but they are acquired by learning.

Notes References

20211217203347 Intelligent agent

20211229192654 Takeaways from animal behaviors applied to robotics

20211217203220 INDEX - Artificial Intelligence

References

(Murphy 2000)

Murphy, Robin. 2000. Introduction to AI Robotics. Intelligent Robotics and Autonomous Agents. Cambridge, Mass: MIT Press.