5.6.2 Storing perceptual predictions and future-action sequences

In preceding chapters we have discussed the role of predicted perception, and of generated future action sequences. In this chapter we discuss how these may be stored. It should be clear that they have to be stored, even if for relatively short periods of time.

We have only discussed memory as storage mechanism for past perceptions and actions and for present perceptions and actions being recorded. Since predicted perceptions and generated action-sequences are most likely based on elements retrieved from memory, we hypothesize that the storage of future-oriented elements uses essentially the same information representation and the same structure.

• 5.6.2 Diagram 1: Adding prediction and planning as extensions of memory
• 5.6.2 Diagram 2: Prediction and planning as part of memory

Let us start by exploring some similarities and differences

• In the past and present, actions and perceptions are linked synchronously and are linked in a continuous time-sequence
• In our simple model we are assuming that there are no memory gaps, i.e. that both perception and action are recorded continuously.
• There cannot be memory gaps on the action-sequence recording side since action-sequences are replayed exactly, muscle twitch by muscle twitch
• There could be gaps on the perception side as long as there are enough to identify the relevant context for action sequences or adaptations.
• For the future, actions and perceptions are not linked synchronously.
• Actions have to be in a continuous time-sequence for our 'raw' information model.
• Single images may be sufficient for the predicted perception (e.g. the place where the predator catches the prey).
• The 'generated' action is likely composed of basic action components from memory, just put together in a different way
• Totally novel movements that are not imitated and that have never been done before are probably very rare.
• The 'predicted perception' image is likely composed of basic perception blobs and background from memory, just put together in different ways
• Totally novel images that have never been seen before are probably very rare.

Given the similarities in content and basic structure, we would hypothesize that information for the future has the same representation as information for the present or for the past.

• Look for relevant experimentation in the literature!