The idea is put forward that the biological brain is full of little predictors/learning agents in the form of individual neurons and their associated synapses. I suppose that is not exactly equivalent to vector based associative memory and yet they are quite similar.
Anyway it is only a matter of a year or two until someone eventually fuses predictors/ associative memory with current deep learning technology in a satisfactory way.
I would say robots like these fictional ones are plausible in the near future and in fact greater sophistication seems likely to be obtainable.
Of course, hopefully they will be better behaved.