This is our ultimate implementation for the topoiesis framework [More]. The sHerazade system adds autonomy to the components, achieving some measure of self-awareness. Ambiguity, analogy and quantum interaction are supported.
Topoiesis is a general logical framework that moves situation theoric reasoning into categoric abstractions so that robust, realistic and implementable models result. Several implementations of topoiesis have been studied, based on three decades of targeted research, funded in early states by US Government funds.
The virtual enterprise domain [More] maps infons (an infon being a fact, situation or reactive predicate) to streamed functions to aggregate and operate groups of collaborative processes among several companies.The intelligence workstation [More] focuses on soft (that is, human, analogical and non-monotonic) reasoning and an introspective distributed web-based collaborative federation. The eidetic streamer [More] deals with a quite different engineering problem dealing with highly specialized recognition services collaborating over massive numbers of streams.
All of these build towards an ultimate system, which we call sHeherazade, or simply sHe. The topoiesis model for sHe has each infon at all levels of granularity not only a process model, but an executable model — effectively an agent in a non-linear self-organizing system.
This is a new approach towards the traditional goals of artificial intelligence, but with an important difference. Rather than achieving reasoning through cognition per se, we use the complexities, tensions and contradictions of stories to make (and re-make) sense. sHe's streaming narratives undergo endless revision, using narrative tension, conflicting information and conditional truths that elude the current state of the art.
The design is notional, and not yet prototyped. However, all the decisions in the basic logical framework (topoiesis) and foundational work in terms of implementation (intelligence workstation network, virtual enterprise integration framework, multistream collaborative object recognition, graphical concept models) have this end goal in mind.