Weird how we’ve been doing this for so many years! I look forward to what you have to say. Dunno if you’ve seen this, may be of interest.
We built some stuff similar to logic graphs, we called distinction networks (d-nets), in the deep past. Here’s some implementation details (1995) for asynchronous d-net computation. Ran it first on an Intel Hypercube with 16 nodes (ugh, course-grain parallelism — a technical abstract (1987) at “The Losp Parallel Deduction Engine” (PDF)) and eventually migrated to a distributed network architecture in which each node was an independent operating system, more for the convenience of doing VR than for the elegance of fine-grain logic parallelism.
Abstract. Intelligent systems can be modeled by organizationally closed networks of interacting agents. An interesting step in the evolution from agents to systems of agents is to approach logic itself as a system of autonomous elementary processes called distinctions. Distinction networks are directed acyclic graphs in which links represent logical implication and nodes are autonomous agents which act in response to changes in their local environment of connectivity. Asynchronous communication of local decisions produces global computational results without global coordination. Biological/environmental programming uses environmental semantics, spatial syntax, and boundary transformation to produce strongly parallel logical deduction.
- Bricken, W. (July 1995), “Distinction Networks” (PDF).
Thanks for the readings. Maybe I’ve just got McCulloch on the brain right now but the things I’m reading in several groups lately keep flashing me back to themes from his work. What you wrote on distinction networks took me back to the beginnings of my interest in AI, especially as approached from logical directions. There’s a couple of posts on my blog where I made an effort to point up what I regard as critical issues. I’ll reshare those next and see if I can throw more light on what’s at stake.