Synthetic Philosophy and Deductive Engineering
Without claiming a broader significance, I think of certain ideas as seminal insights which have inspired and shaped the design of the SPaDE project. This document is an attempt to gather those insights together, in a very concise way, with links to more detailed discussions where available.
The idea that there are certain foundational institutions which are universal for the representation of declarative knowledge.
That concrete syntax is important primarily for communication between humans and machines, and between humans, but that for the purposes of representation within a machine, and the intellectual processes of machine intelligence, concrete syntax is dispensable, and that all knowledge can be represented in a single abstract syntax.
The idea that perfect information spaces and formal deductive theories are equivalent, and hence can be automated by similar methods, notably by self-exploration of the space without need of training data. This connection suggests that generic support for perfect information spaces (which have typically separately engineered solutions for each space) can be provided by a deductive kernel supporting formal theories. The general framework established by a formal foundational institution (such as higher order logic) can be specialised to particular perfect information spaces by defining that space as a formal theory within the foundational institution, and then building the heuristics to solve problems in that space using neural nets focussed on that application.
That focal methods exploiting perfect information spaces can achieve competence in domains which are currently considered the province of large language models and other statistical AI methods. That the hierarchy of logical contexts in the SPaDE knowledge repository can be exploited by focal methods to achieve competence in a wide variety of domains. That singular foci an be identified in multiple domains of competence exploiting reflexive capabilities to accelerate learning by application of intelligence in the domain to the learning process itself.
Two related ideas about evolution:
That formal deduction should subsume computation as the principal paradigm for information processing, and that this shift can be effected by the application of artificial intelligence to the management of the complexity of formal deduction.