CM2001 Article: Gary Banham

To begin setting out the argument that there is an intimate and creative connection between the project of artificial life and new interpretations of transcendental philosophy it will first be necessary to provide as clear an account as possible of what the research project of artificial life involves. This area of enquiry is still in the process of development and there are considerably rich debates between researchers as to the precise outlines of their province. However, Christopher Langton, the organizer of a series of conferences on this topic has provided an account of the general area both in a “preface” to the first volume of proceedings and in his own contribution to these proceedings. Since the former includes a standard baseline account of the area in general I will begin with this before progressing to look at Langton’s overall account as provided in his own article. Whilst this will limit my account to a mere overview of the area and leave aside the difficulties involved in particular projects this approach will have the advantage of a high level of generality thus permitting easier access to the philosophical underpinnings of the whole research programme.

In the “preface” to the first volume of proceedings Langton describes what he terms “the distilled ‘essence’ of artificial life” in the following manner:

There are a number of important points within this short paragraph. Artificial Life is here understood as revealed to be present through the manifestation of certain types of behaviour pattern characteristic of living things. The cardinal distinction between local interactions and global connections allows for an extrapolation that there are occurring developments of a type whose pattern we have given in embryos. Structures are taken to involve a high degree of order that requires support from lower-level entities, a support that is only available to those “life-forms” able to compete effectively for it. “Emergent” structures govern the development of the lower-level entities by establishing a context within which they develop local rules and hence such rules come to “evolve”. Since the “life” in question has to be man-made to count as artificial and since it is determined as “life” because there are “no rules for the behavior of the population at the global level” which latter condition indicates that there is openness in development which comes from self-organization we can see that this description marries one of the central tendencies within contemporary evolutionary theory to a set of machines whose existence is hence circumscribed as involving more than simply use by humans.

Langton offers an alternative and much briefer description of the research programme in his article contributed to the workshop:

The alteration of focus in this definition from the previous description of the “essence” of artificial life is instructive. Whilst the former description utilized a sophisticated vocabulary of levels to orchestrate an understanding of the internal characteristics of the “life” form in question this definition simply relies on the behavioural criteria to state that the closeness of this phenomena to a “natural” one enables us to speak of life. The key difference however is that this second statement of Langton’s clearly explicates the connection he sees between this research programme and the traditional activity of biological science. Whereas biology is stated simply to look at the inner workings of organisms this research programme will demonstrate how to initiate the workings of organisms through constructing them.

The presentation of a project of initiating the creation of life requires a very clear view of how the properties of “life” relate to the manifestations these have taken to date. Hence Langton states clearly that the research undertaken under the rubric of AL views life 'as a property of the organization of matter, rather than a property of the matter which is so organized' (1988:2). Hence life is given a formal rather than a material basis and the latter is comprehended as the province which biology has hitherto been concerned with. Whilst this is one way in which the distinction between an “analytical” and a “synthetic” approach is to be understood it is not the most important one that Langton provides. He adds that the approach of biology has been to view the organism as a whole entity and then to work downwards to an appreciation of its components whereas Artificial Life will instead start 'at the bottom, viewing an organism as a large population of simple machines' and work upwards (which is taken to be a “synthetic” method). This leads to a stark statement of the nature of how behaviours arise and what in sum “life” itself is taken to consist in:

The distinction between “synthetic” and “analytical” approaches is not drawn in a manner that is traditional. A “synthetic” approach is understood by Langton to involve subscription to a basic atomist presupposition that takes the elementary units of a form to be what basically produces the possibilities of that form. The departure from a strict atomistic approach undertaken through the reliance on the notion of “emergence” indicates however that whilst 'each part is a behavior itself' that it is only the interaction between the parts that involves the production of “life”. Hence “life” itself emerges from a set of parts which themselves are each to be considered as active elements yet which only in combination are capable of performing the synthetic operation which gives us “lifelike” behaviours.

The first question that arises about the endeavour to create artificial life therefore is how a set of elements which are not themselves truly “alive” can in combination be said to constitute a “life”. The rationale for thinking this possible emerges from both the understanding of evolutionary processes and the new type of comprehension of “machines”. With regard to the connection between the two we come upon the important notion of “program”. Since the term “program” designates a control structure what is important about it is that this control structure is “abstract” in its operation, that is, it can be stated in a set of rules 'without regard to the material out of which the machine was constructed' (Langton, 1988:11). Hence the basis of the machine is found not to rely on any particular components but rather in the code that is transmitted through these components. As Langton puts it:

If the machine is to be understood formally then it is reducible to the terms of a logic and this logic if stated in a number of distinct but formally equivalent ways is capable of running many different types of operation with a mere variation in the statement of the code. Since the reduction of the machine to its logic is to understand machinery itself as simply a manifestation of coded statements in particular formats then the basic research programme of Artificial Life simply states an extension of the properties of machinery to life itself, understanding the latter as a set of coded elements in particular combinations. That this programme relies therefore on the work of Watson and Crick in terms of a new specification of living properties is clear but perhaps less clear is the extent to which the work of the latter already itself was based on translating the properties of “life” into those of machines. Since the notion of “program” had already been applied to genetic material from the comprehension of computability as a general mechanism for the realization of code, life was comprehended in machinic terms, an achievement which then allows for the understanding of machinery itself as being a “life-form”.

In order for this conceptual revolution to be plausible it is first necessary to re-present the theory of genetics in terms of a generalizable set of statements. This is done through exemplifying the notions of phenotype and genotype as notions permitting a coding which is not specific to any material. Langton hence speaks of a GTYPE (generalized genotype) and a PTYPE (generalized phenotype) which are then described in terms of a recursive logic: 'the local developmental rules - the recursive description itself - constitute the GTYPE, and the developing structure - the recursively generated object or behavior itself - constitutes the PTYPE' (1988:25). Since by this process a computer can simulate the characteristics of any machine we wish to instantiate it becomes possible to present complex behaviour based on simple rules and to see the rules as key to the behaviours we term “life”.

Whilst Langton’s account of Artificial Life may not be in all respects equivalent to that adopted by others within the research programme its general parameters do not seem to be disputed. It will hence be relied on in the account I now wish to give of a convergence between the research programme and some leading interpretations of transcendental philosophy.

The notion of Artificial Life is a development emergent from the research programme of Artificial Intelligence.What I wish to make clear in considering a particular line of interpretation of transcendental philosophy is how the gap between the notions of Artificial Life and Artificial Intelligence can be reconceived. This form of interpretation of transcendental philosophy will turn on showing how a functionalist theory of mind can be presented not as an empirical “discovery” but rather as a set of conditions for the possibility of representation occurring for a “mind” at all. Reconceived in this way the functionalist account of mind will be thus conceived as the “inner truth” of transcendental philosophy.

Presenting this interpretation of transcendental philosophy will require drawing upon some contemporary readings of Kant. The reading of Kant I will be exploring is provided by Andrew Brook but before referring to his work in some detail and drawing out the sense in which it provides us with a linkage between intelligence and life in a manner which enables the gap between the programmes of AI and AL to be considerably narrowed it will first be necessary to underline the limits of Brook’s account and the aspects of it which I will not be accepting. Brook considerably weakens his interpretation by presenting it as based on an empiricist notion of mental functioning which he opposes to the “Cartesian” picture of mind which he imagines to be entailed by transcendental idealism. Since it is a natural error for an empiricist to suppose that someone whose position is divergent from his own is a rationalist perhaps this notion of a “Cartesian” transcendental idealism should not be entirely surprising. Since it supports a radical failure on Brook’s part to investigate the notion of “transcendental” beyond the level of imagining that there are “transcendental arguments” separable from any positive notion of philosophy the full possibilities of a functionalist form of transcendental philosophy can not be exploited by the author who is primarily responsible for alerting us to them.¹

With these very important caveats made it is plausible to construe the interpretation presented by Andrew Brook as providing us with a notion of transcendental functionalism which will enable us to see how the programs of AI and AL can be brought together into a coherent whole. The key to Brook’s interpretation is what he terms “apperceptive self-awareness” (ASA) and, in examining the notion of it he takes Kant to have, he states:

Beginning from this notion of a reference which involves no ascription of properties enables Brook to suggest that at the base of information-processing is a broad and anonymous operation which does not require in any “thick” sense a consciousness of self. How then does this nonascriptive awareness arise? This is equivalent as a question for Brook as to how it is possible to generate a “base” representation of ASA. Brook indicates that this “base” is provided by an act of “transcendental apperception” that generates a “global representation”. Since the “global representation” is the state of intentional directedness to a particular object conceived as the object that it is, it begins to appear that Brook’s reconstruction of Kant’s argument suggests that intentional relation to objects of representation is dependent upon (and in some way brought about by) this “transcendental apperception”, which is also the root of nonascriptive awareness of oneself as oneself. Hence there is a root to both nonascriptive self-awareness and intentional relation to objects considered as unitary items of experience.

How is this common root to be described and what entitles it to be presented as the unique basis of information processing? The first point Brook makes is that apperceptive self-awareness is distinct from most forms of self-awareness such as looking at oneself in the mirror due to the fact that ASA involves being aware of oneself as a subject, not an object. Since Brook further interprets “object” as “intentional object” this commits him to the view that for Kant there is a non-intentional relation which is the basis of intentionality and hence the basis of the possibility of mental processing in general including the self-ascriptive senses of “subjectivity” that we usually assume to be the ground of self-reference. This is explicated by Brook in the following manner:

So if the core of apperceptive self-awareness is given in the transcendental unity of apperception and if the transcendental unity of apperception is also the basis for the synthesis of recognition in a concept that enables us to apply a rule of description to one thing through time and describe it as the same thing then there is an intrinsic interconnection between non-ascriptive self-awareness and object identification.

This description of mental processing has as its direct consequence the result that self-identification is equated with the awareness of synthetic operations. As Brook writes: 'I am aware of acts of TA [transcendental apperception] and of myself as their subject not by sensing them as the object of a representation but by doing them' (Brook, 1994:80). Hence the distinction between the base of ASA as given in TA and the presentation of identification of intentional states is that whereas in the former there is only awareness in the sense of action being understood to be taking place within the mind, in the latter there is an embedded difference between the intentional object and the awareness of an ability to relate to the intentional object as an object (as it has to become an object for me and hence is distinct from me).

In order to fill in how this model gives Brook an account of information processing that allows a neutrality of material to be presupposed and hence permits a formality of comprehension which can allow us to postulate conditions in general of thinking we need to turn to a set of distinctions he makes between different forms of mental states, distinctions which in turn guide his comprehension of Kant’s transcendental philosophy. Brook sets out a difference between four different types of awareness. “Awareness^I” or “awareness (information processing)” is naturally the most basic notion Brook uses and he describes it as involving:

Whilst this is an intuitive base notion of awareness it alone seems very insufficient as whilst it gives some sense to a relation to intentional objects taking place in a set of processes it does not give us much comprehension of something acting for anything as an intentional object. To fill out the intentional nature of the intentional object further therefore Brook supplements the notion of “awareness^I” with a higher level of “awareness”, namely “awareness^R” or “awareness, representation” which notion he fills out in the following manner:

Whilst this notion of “awareness” is clearly circular it captures a notion of “awareness” much fuller than “awareness^I ” and with the second condition we have given a sense that the object of which one is “aware” takes its properties from one’s type of awareness not from some independent state.This hence renders the object in question “intentional” rather than merely “objective” and in so doing makes it an item of “awareness”.

That this still is insufficient for a full account of “awareness” seems to Brook clear however in the sense that whilst we now have a comprehension of how an intentional object is represented to us it doesn’t involve an account of how reflective relationships to intentional objects can take place such that we can discover that the intentional object before the mind now is precisely the one necessary in order for us to achieve some set aim or purpose. The lack of ability of the two given notions of awareness to capture this ensures that a crucial element of intentionality is not yet given. Hence Brook sets out the notion of “awareness^RR” or “awareness, representation, recognition” which is described as: 'being aware of X, plus recognizing X'. (Brook, 1994: 50). The difficulty with this notion, a difficulty very quickly acknowledged by Brook, is that it incorporates disparate types of “awareness” and that there is clear value in distinguishing between these. This constitutes a potential challenge to the third type of synthesis traced by Kant in the Transcendental Deduction, the synthesis of recognition in a concept. Brook suggests that there is a difference between recognition in a concept and recognition in consciousness and describes the former as “awareness^RR-CC” and the latter as “awareness^RR-CS”. In the former type of awareness we merely have the ability to apply one or more concepts to a representation whilst in the latter this ability has to be something that is reportable and this report indicates the recognition by the one making the report that they are making it. This last form of awareness is intimately related to the type of awareness involved when a recognition of something occurring is only possible because it is occurring, a state entitled by Brook “awareness^H” and defined by him as: 'recognizing a representational state one is having on the basis of having it'. (Brook, 1994: 53) This finally brings us to the base notion which Brook has been seeking to justify, namely the situation whereby the relation between representer and represented is recognized as such and is, as such, the basis of the representation then being given. This is what Brook means when he states: 'when one is aware not just of the object and oneself but of representing the object, it would seem that this representation is the representational base of all three acts of recognition, all three acts of awareness^RR' (Brook, 1994: 81).

The awareness of oneself as subject is thus not a form of awareness like the other types traced as there is no “object” of awareness in this self-construal. There are not properties of oneself being recognised in the awareness that involves the recognition that one is aware of having awareness. 'It follows that, so far as anything my awareness of myself as subject could tell me, I could be any object or any compilation of objects or any succession of objects whatsoever.' (Brook, 1994: 90) The importance of this result as a reading of Kant now needs to be drawn out and its relevance for both the notion of Artificial Life and the relationship between Artificial Life and Artificial Intelligence made clear.

The result of Brook’s reading is that it is both possible and desirable to undertake a discussion of Kant’s philosophy that interprets it as presenting a functionalist account of mind. This possibility centres on the important double role of the transcendental unity of apperception. Whilst it is presented by Kant as 'the supreme principle of all use of the understanding' (B136) there is also a genetic approach to its role in the mind provided within the Transcendental Deduction that shows it arises from other mental activities. Traditionally this argument is stated to have been made in the A-Deduction and abandoned in the B-Deduction. However, whilst the A-Deduction provides us with reasons for seeing the transcendental unity of apperception as based upon the activity of imagination [2] what Brook’s account makes clear is that there is a “subjective deduction” in the second edition of the First Critique, a “subjective deduction” that makes the analytical priority of types of awareness that are not implicated in self-awareness talk over those types of awareness that are implicated in self-awareness talk clear.

Not only is this an interesting result for Kant experts but it has strong implications for a relationship between a “Kantian” theory of mind and the type of explanation we have seen given in Langton of how an artificial life is organized. The proto-functionalist account of Kant brings out a mutual relationship between the “synthetic” operation of levelling down to the base of “information processing” and levelling “up” to the opposite base of ASA. The two types of base of mental functioning are co-dependent and co-producing which makes the premium placed upon the notion of consciousness as the action of an agent hard to sustain. Hence the operations performed within a mind according to this reading of Kant are similar in kind to those that would be performed in life by an artificial organism. Just as the artificial organism could not relate to its environment except through an emergent logic so the Kantian “machinic” intelligence requires a principle of supervenience for organisation to take place. This crucial connection also makes clear in what a link between artificial life and artificial intelligence would consist.

Both AL and AI require a reference to action that arises not from a primary centre to the periphery but the converse, the organisation of a centre from the inter-locking emergent behaviour of the parts. This dependence of artificial forms upon a “bottom-up” approach does not prevent them from claiming a reference to life or intelligence if it is the case that intelligent life forms already require such a directional relation in order to function at all efficiently. The connections therefore between a proto-functionalist account of transcendental philosophy and the research programmes of artificial life and artificial intelligence become clear. In both cases there is a relationship between a highly general description that requires no reference to the matter organized and an intricate dependence upon functionalist inter-connection. In both cases this has the result of formulating a set of general conditions of possibility of intelligent life beyond the specific conditions of humanity.

Towards the close of the treatment of the Transcendental Deduction in the second edition of the Critique of Pure Reason Kant writes:

Concepts of the understanding are justified immanently. Kant, in closing the First Critique, further argues that it is additionally the case that reason is immanently self-developing (A835/B863). This suggestion of a necessary immanence of rational principles and laws suggests that for cognition to be taking place there must be a relation between the elements of cognition and its overall grasp. This is an integral relationship, one described by Kant in a number of places as “purposive”. Kant took it be the case that such “purposive” relations where not available to what he termed “machines” and this is an important element of the Critique of Teleological Judgment. The notion of “machine” with which Kant was dealing was however one modelled on Newtonian notions of physics and related to a network of efficient causes.

The problem with the conception underlying the research programme of Artificial Life however is that the notion of “synthetic” organization within it flouts the fundamental principle of teleological connection expounded by Kant. The reason for this is not primarily because of the fact that the “life” being spoken of is post-organic but that the principles governing the organization of the “life” in question are bound to Leibnizian principles of relation and basically function as monads. The monadic organization is said to produce an emergent logical progression that depends upon an assumption of a pre-established harmony between the parts given.

Self-replication is only one of the principles of life-like organization. Connected to it is a recursive connection between the parts organized, a connection requiring that the elements of the whole are effected by the organization they constitute in addition to forming such a whole. Whilst a recursive connection between the GTYPE and the PTYPE is argued to be logically necessary by Langton the relationship between this recursive logic and the emergent notion of organization is unclear. Furthermore, in relation to the linkage suggested between the reading of transcendental philosophy thus far presented and AL it is worth revisiting the structure of Kant’s Critique of Pure Reason from a different angle. Thus far we have presented only a logical reconstruction of the transcendental deduction in which the B-Deduction was patterned on the A-Deduction. The recursive nature of reason is however insisted upon by Kant in the “Appendix” to the Transcendental Dialectic:

The possibility of the operations of the discursive understanding is provided to it by a rule of reason. This rule indicates that the possibility of progression in comprehension occurs through the following of the procedure of unificatory synthesis. As Kant adds: 'For the law of reason to seek unity is necessary, since without it we would have no reason, and without that, no coherent use of the understanding, and, lacking that, no sufficient mark of empirical truth; thus in regard to the latter we simply have to presuppose the systematic unity of nature as objectively valid and necessary' (A651/B679). From the notion of systematic unity we derive logic itself, both in terms of such key formulations as Occam’s Razor (A652/B680) and the possibility of applying logic to nature at all (A654/B682).

The principle of all principles is purposiveness itself. Without this principle that guides and directs the postulation of all other principles of investigation we would have no possibility of setting out to examine with precision the relations between phenomena. However, not only is the purposive unity postulated itself 'a mere idea' (A681/B709) but it is also subject to continuous modification as the discovery of new connections through the divisive operation of the understanding requires this. The nature of this constant revision indicates that the notion of purposive unity is a practical guide to investigation not an object of theoretical knowledge.

This is the point at which the conception of Artificial Life presently given is shown to be problematic. Since there is given in its conception only an emergent logic the operation of the life-form does not permit cognition in the ultimate sense of grasping the notion of law-like connection and nor is the “life-form” in question itself governed by such connection. This failure of harmony between the Artificial Life and the organic is decisive for the mechanistic basis of the former and prevents it from being cognisable as operated according to transcendental principles.

Recursive operations are said to operate only once the level of complexity of the “life-form” given in Artificial Life experiments has sufficiently “emerged” whilst transcendental accounts require this complexity to make possible the notion of “awareness” in general. Hence Artificial Life forms lack epigenetic organization and in fact have the status of “machine” in Kantian terms not because of the stuff from which they are made but because of the logical interactions which govern their operation and their possibility of comprehension. This also indicates the crucial link between AI and AL as being mutually harmful in cognitive principles.

Whilst the above considerations suggest a lack of future for the research programme of Artificial Life there is another way of addressing the nature of the project involved which indicates quite contrary conclusions and enables a final harmonization between the outlines of this area and a machinic notion of transcendental philosophy. This is provided from the work of Bernard Stiegler. Working from across the divide between philosophy and anthropology, Stiegler has presented some original reflections on the nature of technicity. Writing about the nature of tools Stiegler refers to their capacity to “store” information that transcends the life-spans of organic individuals. The capacity of a storing system that survives those who have apparently “invented” it to be accumulatively invested in by future generations requires a reflection on the relation between organic and non-organic forms.

According to this analysis the emergence of recursivity is given in the relation between the organic and the technical. The relation between them is one of interlocking and mutually interactive development in which the progressive organization of a gigantic “memory” takes place. New memorizations facilitate that which they have anticipated and life is available to be seen as part of a vast enterprise of “programmability”.

If the relation between organic and inorganic is what is recursive and the self-relation of the organism is constituted through technical supplementarity then there is in principle no problem with viewing “artificial life” as coterminuous with culture itself. This was also the conclusion of Kant’s Critique of Teleological Judgment: 'Producing in a rational being an aptitude for purposes generally (hence [in a way that leaves] that being free) is culture' (Ak. 5: 431). Culture is the capacity to act and think purposively, being-in-community is to be able to work, that is, to be free. But this development is a product of what Hegelians (following here Kant) term “second nature” or artificial life. The increasing complexification of the outward manifestation of the gigantic “memories” which make possible activities of a type which are responsible for futurity just is the process of reaching for the self-conscious programme of conceiving of programmability itself as the development of life.

The development of programmability as an end-in-itself is the self-surpassing of the human on the basis of its own premises of development. Put within this larger context the surprisingly limited conceptions of artificial life to date are merely a consequence of failure to reflect on the conditions of possibility of being able to theorize on artificiality at all. Reaching the horizon of reflection which permits this is to see that the transcendental nature of organicity permits and enforces an “artificial” environment as the ability to plan is part of an anticipatory systematisation that lacks final horizon (is “infinitely open”). This way of phrasing the situation of the arrival of the post-human future indicates that the human has never been one with itself, is that which is incapable of being “at-home” and whose historicity is never past but futural.

Given the transcendental perspective of purposiveness we can now see that whilst the thought of teleological connection indicates a problem with seeing the research programme of artificial life as being concerned truly with life this is purely because of the self-limitation of enquiry which this programme has allowed itself to be given. If Artificial Life is taken as the futural name of transcendental philosophy then the way is open for conceiving its products and programme as nothing other than the search of the future for itself.

[1] It is also worth mentioning at least once that the notion of “mind” is a weak translation of Kant’s gemut. For an account of the full complexities of this term in Kant see Howard Caygill (1995) A Kant Dictionary. Oxford: Blackwell.

[2] For a fuller account of this argument see Gary Banham (2000) Kant and the Ends of Aesthetics. London: Macmillan & New York: St. Martin’s Press, Chapter 2.

Langton, C. (ed.) (1988) Artificial Life: The Proceedings of An Interdisciplinary Workshop on the Synthesis and Simulation of Living Systems Held September 1987 In Los Alamos, New Mexico. California: Addison-Wesley Publishing Company, Inc: Redwood City.