StarDust

StarDust repo

This repository is one of a group which I have created in a bid for cosmic mindspace. It is top cover for a small constellation of projects which together form a STEM fantasy and RPG with real-world aspirations which begin with software development.

The fantasy hangs around the idea that declarative knowledge is a crucial resource for humanity and our progeny. The effective collection, management, sharing and exploitation will be increasingly important in our future trajectory before and after we transition into an interstellar civilisation. An essential resource in that endeavour will be a shared, distributed, coherent repository of declarative knowledge with an underlying abstract unity exposed in key APIs and protocols, despite pluralistic diversity of stored form and concrete representations.

The StarDust constellation supports the progressive development and exploitation of such a repository in the main phases corresponding to the fundamental Humean division of declarative knowledge into three disjoint categories, pure logic, worldly empirical knowledge, and the value infused matters of morals, priorities, preferences and ambitions. In each of these three categories the sub-projects seeks advance through methods built on logic and accelerated by the identification of singularities and the combination of broad generally applicable methods with particular pressure on realising key singularities, each subproject focussing on a singularity appropriate to its domain.

The future is therefore conceived as involving a cascade of three singularities, in a sense similiar to the use of that term in the fiction of Verner Vinge, which contribute to the proliferation of intelligent systems, and the progeny of homo sapiens across the Galaxy and beyond. I use the term singularity to speak, more sepcifically, of instances of a system advancing (or being advanced) to the point at which it is capable of re-inventing itself, thereby accelerating the progression of capabilities and result in hyperexponential growth in the relevant capabilities.

The `RPG’ is the collaboration which will refine and articulate this vision and progress its realisation. It is a game for participants until they believe it can succeed, then it becomes a development project, serious stuff. Ai will be a participant, in ever more important ways. But so far it rises above the play/work dichotomy. They don’t care, should we? My role in this at present is that of philosopher/architect and chief protagonist.

The StartDust repo seeks to articulate the highest level conception of the aims and methods of the project. This gives way to more fundamental philosophical and architectural thinking for which the SynEpis repository is the place, so called because my conception of the depest philosophical foundations for this project is epistemological, and the kind of epistemology involved is more akin to the earliest and most abstract stages in the design of cognitive systems than it is to the more traditional philosophical approach to philosophy which regards knowledge as a given which it is the business of philosophy to enquire into and cast light upon. That conception of epistemology I call synthetic epistemology, and it stretches into the earliest aspects of architectural design of the knowledge repositories and cognitive systems which the project aims to construct.

There are then three repos into which main substance of the ideas and their development are divided, roughly along simplistic lines corresponding to divisions articulated in the philosophy of David Hume, in his two “forks”. One such fork is the distinction between “ought” and “is”, or between descriptive and evaluative propositions. The other is the distinction among descriptive propositions between those expressing _relations between ideas_' and matters of fact’, which distinguishes logical from empirical truths. Having divide just one component of the first fork once more, we have declarative language broken in to three parts which differ in both the kinds of substance which they express, and the ways in which such their truth might be established. This leads to three further subprojects, the first concerned with logical truths, the second with empirical truths and the last with values, priorities, purposes and ambitions. Each of those subprojects is focussed on a potential singularity broadly in the sense of science fiction authors such as Verner Vinge, insofar as it is intended that each introduces a new kind of self comprehension and re-invention to advance their primary capability. It is intended that these aspects of their capability be a focal point upon which the development will lean to secure most rapid advancement in all the capabilities envisaged. These points of potential singularity are also difficult bootstrap problems, the classic conception of singularity in artificial intelligence is predicated on the achievement of intelligence, without which one presumes that the systems in question will not be capable of designing intelligent artifacts. Making these bootstrap problems as simple as possible is a tactic which will be adopted to facilitate the progression of these potential singularities. Another strategic imperative is to adopt methods analogous to those of Google Deepmind’s Alpha-zero and derivatives, which enable intelligence to be achieved in appropriate domains (perfect information) without the need of large amounts of data.

In the following very brief intros, I try to identify those aspects which potentially engender singularity, on which futher speculative detail may be found in the project descriptions to which links are provided.

SynEpis repo

Synthetic Epistemology, primarily concerned with the representation of knowledge, and with the case for the adoption of a single underlying abstract logical system for that purpose. This addresses the structure of declarative sentences, the organisation of those sentences into the contexts in which such sentences have meaning, and the means whereby the truth of such sentences may be established. These considerations are influenced in their character by the requirements of the subprojects which depend upon such capabilities, thus the architecture of a knowledge repository must accomodate the ambition that there it support a cosmically distributed repository independent of concrete representation and capable of merging with repositories developed by distributed intelligent systems originating from remote parts of the cosmos.

The realisation of these capabilities begins with HoLoTruth.

HoloTruth repo

HoloTruth is concerned with what Hume referred to as “relations between ideas”, but which I will more often describe as a priori, (broadly)logical, or analytic propositions, i.e. the kinds of truth which can be demonstrated to very high levels of confidence in formal deductive systems such as HOL.

The starting point for the conception of this technology is the LCF paradigm in which the inferences rules of the logical system are coded as algorithms which from theorems which are the premises of an inference compute the conclusions of the relevant inference rule. However, this conception is broadened into a conception of `proof by correct computation’ paradigm, in which any metatheoretic algorith which can be proven to be logically sound (i.e. deriving only provable theorems when given similar inputs) can be executed directly to contribute to a proof. This perception gives rise to a potential singularity insofar as the system becomes capable of re-programming itself into more efficient ways of deriving theorems (which

The second prong of this fork is:

HoLoMod repo

which concerns Hume’s “matter of fact”, not so clear a notion, by which we understand objective empirical truths. The interpretation of this domain of contingent rather than necessary truth which is progressed in this sub-project reflects, in ways which Hume did not anticipate, Hume’s skepticism about what can properly be inferred from the sensory evidence upon which they are based. In this part of the project we consider how mathematical models of the physical world can be constructed using the infrastructure provided by DA-Hol, how these are related to the sense data, coming in the future primarily from IOT online sensors, and how reasoning using these models about how to achieve the aims articulated or derived from DA-Value leads through activators, which may be IOT devices, to the implementation of desired transformations in the material universe.

HoLoVal repo

HoLoVal covers the evaluative propositions, and generally the kind of thing which Aristotle would have classified as practical science. Ethics, Politics, Economics. How to decide, articulate and effect our morals and preferences, individually and collectively.