Synthetic Philosophy and Deductive Engineering
In the SPaDE project, philosophy and engineering are cheek by jowl, and the distinction between them is not clearly made. The philosophy is synthetic, in the sense that it is constructed to serve a purpose, and the engineering is philosophically grounded and built on those foundations to realise that purpose.
A central point of engagement is through the emphasis in deductive engineering on the exploitation of knowledge in the solution of real world problems, and the importance of epistemology as a fundamental approach to knowledge and the design of infrastructure for the management and exploitation of knowledge. Once philosophy is thought of as synthesis rather than an observation, it becomes a necessary prelude to the architecture of knowledge repositories and application of deductive methods to engineering problems.
Logical Positivism was that manifestation of positivism created in the wake of the great advances in the logical foundations of mathematics which took place around the turn of the 20th century. In the philosophy of Rudolf Carnap, inspired by the work of Bertrand Russell on the formalisation of mathematics and his ideas for “scientific philosophy”, it promoted and facilitated the application of formal deductive systems to the advancement of logical rigour in philosophy and science. Synthetic Philosophy is intended as a successor to Logical Positivism, oriented more broadly to Science, Technology Engineering and Mathematics (STEM) and to all applications in which deduction may play a role.
The structure of this exposition is a bit like a roller coaster. In various ways I think of SPaDE and its philosophy as a stack with certain things at the top and others below. There are multiple presentations having that kind of structure, some more philosophically oriented, in particular the epistemological stack, and some more engineering oriented, notably the focal tower. From the philosophical perspective, at the bottom are the foundations: metaphysics, philosophy of language, epistemology, logic and a priori or analytic truth, in the middle empirical knowledge the STE of STEM, and at the top the third of the Humean trichotomy values, ethics and purpose.
We begin at the top, so that the whole may be considered in the light of the supervening purpose which it is intended to serve, and then look down to the foundations which underpin it, concluding rather more briefly in the middle ground, which for present purposes I characterise as concrete philosophy.
Probably the parts of this philosophy which are closest to academic analytic philosophy are those closest to the philosophical and logical foundations. But these ideas are unlikely to be well understood in isolation from the broader context of the engineering and ethical purpose which motivates them, and determines the synthetic character of the philosophy.
In this document I shall try to address the foundational elements of synthetic philosophy, but feel the need to place those ideas in the context they serve by saying a little about the superstructure which they are intended to support.
Synthetic Philosophy has purpose, and is engineered to advance that purpose.
Synthetic Philosophy is the name I have adopted for a successor to logical positivism oriented more broadly to Science, Technology Engineering and Mathematics (STEM).
That orientation is instrumental. Synthetic philosophy is itself engineered to advance an ethical agenda derived from a rational speculation about the very long term future of intelligence in the cosmos. A sketch of the thinking which has lead to Synthetic Philosophy may be found as a A STEM Fantasy and Ethical Consequence.
It is therefore a purposeful philosophy, anticipating key elements of the longer term proliferation of intelligence with the intent of influencing the moral character of that proliferation. At this point, I have little more to say about that purpose than that the proliferation should be benign, which I associate a preference for cooperation over conflict. Hopefully some elaboration and refinement will emerge as the project progresses, but the project will be mainly focussed on the manner in which its proliferation is engineered.
Synthetic Philosophy has a layered structure which corresponds both to an epistemological stack and a focal tower.
The epistemological stack reflects the tripartite division in declarative knowledge, found in the philosophy of David Hume and also in Logical Positivism, into analytic, synthetic and normative knowledge. This is considered a stack since each layer builds on and makes use of the previous layers. In this, it is similar to a software stack. It is also a foundational structure (of more than three layers) most conspicuous in the lower levels where we find philosophy and logic providing formal foundations for mathematics and all analytic knowledge, which in turn provide a basis for science and engineering. The normative layer at the top of the stack is concerned with values and ethics, and provides the motivation for the whole enterprise of knowledge acquisition and application.
The focal tower reflects the importance of specialisations to the effectiveness of constructive endeavours, and the focus required to address difficult problems in each domain. It recognises a variety of ways in which focus is enabled and exploited, the most basic of which corresponds to the layers of the epistemological stack,
in which abstract deductive reasoning underpins concrete knowledge of the world, which in turn underpins normative knowledge of values and ethics. in which abstract deductive reasoning underpins concrete knowledge of the world, which in turn underpins normative knowledge of values and ethics.
A feature of the SPaDE project is its emphasis on focal methods, which involves the clarification of purpose and the concentration of effort on those aspects of knowledge engineering which are most relevant to the achievement of that purpose.
The SPaDE project is an engineering manifestation of that philosophy, and this presentation of synthetic philosophy is intended to show the main features of the philosophical system in alignment with the goals and structure of the project.
Synthetic Philosophy is primarily epistemological, which is to say that it is intended to underpin the acquisition and application of knowledge. It is presented here as falling into three parts, distinguished by the epistemological status of the propositions which concern them, and the domains of knowledge to which they relate.
These three parts of synthetic philosophy are interrelated. Each is prerequisite to its successors, and a motivator for its predecessors. This relationship is discussed in greater detail as constituting an Epistemological Stack which influences the architecture of the SPaDE project. This epistemological stack of synthetic philosophy is complemented by a parallel structure in the Focal Tower of stages in deductive engineering supporting evolution of information processing to a deductive paradigm.
Though practically, and ethically oriented, synthetic philosophy includes and is constructed upon foundations belonging to what is here labelled abstract philosophy by which I mean:
All these areas are generally construed more broadly than their foundational role in this notion of abstract philosophy requires. It is characteristic of the approach to these matters in synthetic philosophy that they are treated constructively, not primarily as matters to be discovered so much as domains to be synthesised, though a part of the enterprise does involve discovering the consequences of those constructions.
Abstract philosophy is foundational for synthetic philosophy. The great edifice of human knowledge is best thought of as constructed on philosophical foundations, consideration of which has an important role to play in its future advancement as intelligence proliferates beyond its earthly confines.
These deepest philosophical foundations are those parts of knowledge which are least influenced by the particulars of human origins and culture, and among the features of our distant future which we can most reliably extrapolate from contemporary knowledge. The natural numbers, to chose a simple example, are likely to be a part of the knowledge of any intelligent beings, human or artificial, wherever they may be found in the cosmos
In describing these philosophical foundations ab initio, no one of the four disciplines can be independently accounted for, they are intertwined and interdependent, though the emphasis of the SPaDE project on the representation, manipulation and application of declarative knowledge places epistemology as primary among them.
I shall give here the briefest possible sketch of the main features of the four components of theoretical philosophy as they relate to the SPaDE project.
Though there are many kinds of knowledge, the SPaDE project is concerned primarily with declarative knowledge, i.e. knowledge which may be expressed in declarative language, considering this as its most advanced form, to which other forms of knowledge may be reduced.
SPaDE accepts the tripartite division of knowledge into relations of ideas, matters of fact, and value judgements emphasised by David Hume, regarding these as layers in an epistemological stack, which influences the SPaDE strategy for the representation and application of declarative knowledge. Alongside this epistemological stack we describe a more finely grained focal tower which influences the architecture of SPaDE and facilitates the advancement of knowledge through focal engineering methods which focus on the most relevant aspects of knowledge for a given task.
Synthetic philosophy is metaphysically conventionalist. The world is as it is, but in order to speak of we must divide it into parts, and that division is a matter of convention, influenced by our purposes in speaking of it. It is convenient in the establishment of declarative language to think of language in the first instance as consisting of abstract entities conveying knowledge about abstract entities, which may then be applied through the articulation of correspondences between abstract and concrete entities, thereby effecting models which approximate the structure of reality in practically useful ways.
Working in the first instance with such abstract declarative language provides the methods and tools for the construction of models of both physical and ethical systems, and enables reliable deductive methods to be applied in all domains of declarative knowledge.
In order to effect this abstract foundation, it is necessary to chose appropriate abstract domains, and we will find when we come to discuss the expressiveness of declarative languages and the possibility of universal declarative languages, that the key choice to be made in relation to those abstract domains is cardinality. It does not matter what these abstract entities are, only how many there are.
The conventionalist metaphysics of synthetic philosophy may this be reduced to the thesis that we may chose whatever abstract ontology we please, but at bottom the choice is just one of cardinality and the difficult problems in this domain concern how to express the choice of cardinality in declarative language, how to devise propositions which are logically consistent but are only satisfiable in models of sufficient size (in set theory, these are the large cardinal axioms).
These are however, from the perspective of synthetic philosophy, analogous to debating how many angels may dance on the head of a pin, since the practical application of declarative knowledge does not depend on very large cardinals. It is doubtful whether any practical application of declarative knowledge will be expedited by the assumption of more than an unbounded set of inaccessibles. (I will not go into how few might suffice, since practical advantages of ontological parsimony are moot, notwithstanding Occam.)
It is tempting to say that the whole point of declarative knowledge is to enable us to shape our world through the construction of models which enable is to use deduction to predict and control phenomena in the world. But deduction is a feature of declarative language, and we cannot give an account of if without first addressing the nature of declarative language.
The propositions of declarative language express through their truth conditions constraints on how their subject matter may be, which is to say what things the various names in the language may denote. These constraints give rise to a relationship of entailment, which obtains when the constraints imposed by some set of declarative sentences encompass those imposed by another sentence. In such cases we may infer the latter sentence from the former set. This is the essence of deductive reasoning, which is the method by which scientific theories can be applied to predict and control phenomena in the world.
In order to determine when such entailments obtain, we seek an effective method for establishing entailments. For languages of sufficient expressiveness for many practical applications, it is known that no such effective method exists, but practically useful semi-decision procedures are available which suffice for practical applications.
These semi-decision procedures are the subject matter of logic as it relates to declarative knowledge, and the design of such procedures is a key part of the SPaDE project, in which some innovation is anticipated. The procedures are normally defined as the closure of a set of axioms under a set of ‘primitive’ inference rules, and the most rigorous automated support for formal reasoning found in interactive theorem provers (ITPs) are often implemented in accordance with the ‘LCF paradigm’, in which all theorems must be proven by reduction to the primitive inference rules. Increasingly however, ITPs support some kind of ‘reflexive’ metatheoretic extension to the primitive rules in which more efficient forms of reasoning may be shown to be consistent with the primitive rules and may then be applied directly without reduction to the primitive rules.
Such metatheoretic extension is part of a broader application of reflexion forming a key part of the focal engineering methods pervasive in the SPaDE project. A significant instance of this comes in the idea of singular focus, where reflexive methods incorporated into the realisation of some capability enable that capability to improve its own design and implementation.
Concrete philosophy is the domain of empirical truth and physical engineering. In the SPaDE project, concrete philosophy is addressed through the construction and application of abstract models of physical systems, expressed in declarative language and subject to deductive reasoning. We may think of these models as isomorphic to aspects of the material world and given semantically by a bijection between the abstract entities of the model and the concrete entities of the physical system being modelled.
The normative part of synthetic philosophy is concerned with purpose, values and ethics. Why do we need to cover this in an AI dominated engineering enterprise?
This is something which I had not expected to cover until very recently, despite the strong sense of purpose which has driven this project. Hitherto I have thought in terms of evolutionary necessity. The necessity which I have characterised as arising from an evolutionary imperative guaranteeing that the dominant intelligent systems in the cosmos will be those which are best engineered to proliferate, together with the belief that the main fuel for the accelerated proliferation of intelligence will be knowledge.
The project was then conceived as playing into evolutionary necessities, both anticipating and designing the future. An important design strategy was to go for foundational abstractions which owe nothing to the particulars of human origins, of which the simplest example is the natural numbers, but more cogently considering that the adoption of a single underlying representation for declarative knowledge in a universal foundational institution would eventually be adopted by all intelligent systems, human or artificial, wherever they may be found in the cosmos. What then of morality, when the dominant intelligence across the cosmos is shaped by the necessities of transgalactic proliferation?
In two ways I came to see that normative philosophy has a place in synthetic philosophy, beyond being another application of deductive reason.
First I thought of the emergence of morality in the evolution of life on earth. This is a late development, and surely could not have become universal in humans (notwithstanding the diversity of moral systems that have emerged) if it had not been conducive to proliferation. So I reasoned that the same evolutionary imperative which drives the proliferation of intelligence generally would also yield moral constraint, and that a system devoid of moral purpose could not predominate in the long term.
A second consideration, or perhaps a wrinkle on the first, is the consideration that among the various technical developments which are equally conducive to proliferation, would take mindshare which would be advantageous in getting off the ground.
I cannot say much about what core normative principles might be adopted which could persist through the evolution of the next billion years, but a general preference for cooperation over conflict seems to me a contender, and we may hope that this seed might be elaborated and refined as the project progresses.