Common components of other cognitive architectures and the Semantic Pointer Architecture

Question

In the papers I've read about it, the Semantic Pointer Architecture (SPA) embodied in Spaun is said to be more biologically plausible than many other proposed architectures such as the Neural Blackboard Architecture, ACT-R and ICS. However, I was wondering if they had any shared attributes? If this is too broad of a question, which one can SPA considered most similar to?

I am currently reading "How to Build a Brain" by Chris Eliasmith to answer these questions and will post my own answer shortly, but I was hoping for other perspectives.

Answer 1

In "How to Build a Brain", Chris Eliasmith draws parrallels between Spaun and various architectures. All of my quotes in my answer are from that book.

The Dynamic Field Theory of Cognition is similarly pre-occupied with outside stimulation as being essential to understanding cognitive function, consequently "there is little in the SPA that could not be reformulated using terms more familiar to dynamic systems theorists"

Leabra shares an emphasis on implementation on a realistic biological substrate. In particular, the cortex->basal ganglia->thalamus loop for action control/decision, as well as the effects of dopamine on Spaun, were heavily influenced by Leabra.

ACT-R also shares the cortex->basal ganglia->thalamus loop, but more importantly puts a similar emphasis on timing. That is, there is a specific time for switching between tasks and reacting to stimuli that must be followed.

ICS shares the similar mathematical binding operation as SPA, thus many of the results of the ICS are informative for SPA and "how such representations can be related to a wide variety of linguistic processing".

While these similarities give a good reference frame of where Spaun stands in the realm of neural architectures, it is much more interesting (and important!) to note where it deviates. A topic which is obviously covered in great detail in "How to Build a Brain".