2
$\begingroup$

I see people are using a lot the terms sensing, perception, and learning together in the context of artificial intelligence. It seems to me perception is a more board term that may include sensing and learning. What are their definitions and how do they interact with each other? Any paper references are highly appreciated.

$\endgroup$
2
  • $\begingroup$ i could be wrong about this ... perhaps something like an intelligent thermostat in learn mode sense room temperature, perceive that the temperature is dropping, learn that the human will activate the furnace if the temperature drops some more .... sense - read raw data ... perceive - make sense of data ... learn - determine what is the required action based on perception $\endgroup$ – jsotola Nov 8 '20 at 0:19
  • $\begingroup$ My baby robot sensed something hot nearby; he perceived that something was bad, but not very sure; so he used his finger to verify; his finger was burnt: he *learned that next time he should not use his robot finger to verify, perhaps use a thermometer instead. $\endgroup$ – tlfong01 Nov 8 '20 at 1:55
1
$\begingroup$

Sense and perceive, in robotics, are synonyms. There is no practical technological step between sensing and perception. If we take a robot for example, or in intelligent thermostat, the sensor signal is fed to the controller (or agent if we consider reinforcement learning), actions come out of controller (or agent). There is no in-between step practically which would constitute a difference between sensing and perception.

One could argue, that there is a difference between the sensor signal and the robots internal representation of state. If we consider SLAM algorithms or sensor fusion also in this case there is "no extra" step which one could call perception, differentiated from sensing.

Learning is a procedure, which uses data to create models. It is very loosely linked to sensing. Sensing can be used to gather data for learning and sensed data can be passed though a machine learning model to achieve a certain goal (e.g. classifying that signal or identifying an object if the signal is a camera).

The comments to the question are suggesting that perception would be somehow information derived from sensing...That is just signal processing, has been called that for years, the goal of signal processing is not to "create perception" but to process signals and derive other/better information from it.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.