5.3.2 How might basic learning work with 'raw' information?

We start by assuming that we are only dealing with 'raw' information, i.e. as it comes from perception and as it is required for bodily action.

Let us assume that information is stored as it arrives or is used, in temporal sequence.

• This approach has a bias for recent experiences

Successful action presumably leads to a positive feeling such as satiation as indication of a desirable outcome

• Even though we have not included perception of hunger or satiation, let us assume that these feelings are somehow included in the memory sequences

Let us assume we have a negative feeling such as hunger to which can match a specific positive feeling such as satiation

• To repeat a positive feeling, we might simply backtrack through the memory sequence until we find the most recent positive feeling of that type
• We can then backtrack some more, to the beginning of the action that led to the desired outcome.
• To make it more realistic, we might even compare the remembered scene at the beginning of the remembered action to ensure that it resembles the currently perceived scene

To accomplish the goal we 'replay' the memory of the action, to hopefully get the same outcome.

• We are assuming that the previous muscle activity is replayed exactly, by sending the same impulses through the nerves to the muscles.
• This, of course, does not address how new action sequences might be learned or invented.