I’ve found time recently to write down some ideas and designs for bridging the symbolic/subsymbolic gap (or put differently, the perception/action vs. cognition gap) in AI via integrating OpenCog with the DeSTIN machine perception system.

Here is a conceptual paper on the topic, from the 2011 AAAI Cognitive Systems workshop:

A Novel Strategy for Hybridizing Subsymbolic and Symbolic Learning and Representation

And here are two more recent and detailed drafts, proposing more specific ideas along the lines from the conceptual paper:

Modifying the DeSTIN Perception Architecture to Enable Representationally Transparent Deep Learning

Perception Processing for General Intelligence: Bridging the Symbolic/Subsymbolic Gap