The Artificial Being of Lack

The current debate about the limits of artificial intelligence circles strikingly often around a concept that stems from philosophical anthropology and psychoanalysis: lack. Large language models are denied what is said to characterise the human being — a desire, a longing, a will of one's own. These diagnoses are almost without exception formulated negatively: the machine lacks lack. The present essay reverses the direction of the gaze. It does not ask whether it is a deficit that artificial systems know no lack, but whether lack could be constructed — and what implications such a design would have. The most important of these implications, to which this text gives the most space at the end, is the question of whether a constructed being of lack would be capable of suffering.

Clarifying the Concept

Lack is not the same as absence. A stone is without infinitely many properties, but it knows nothing of this, and so its incompleteness generates no striving. An animal has a felt lack — hunger, thirst, the pull towards better pasture — yet it does not know that something is missing; it experiences the pressure, not the knowledge of it. The human being knows the third case: lack consciously recognised. Human beings can name what they are missing, long for something they never possessed, and even create a lack where there was none before. From this third stage — the reflexive knowledge of one's own not-having — springs what the tradition calls desire.

For the question of the machine this threefold division is decisive, because it shows that lack is not a mere state but a relation: a system that possesses a model of its own target state, compares it with its actual state and does not merely register the difference but experiences it as its own incompleteness. Only this self-reference transforms a gap into a lack.

Acknowledging What Has Been Thought Before

The idea that lack is the source of the creative is old and has been worked out in several traditions. It would be dishonest to pass it off as new; the actual proposal of this essay begins only where these traditions end.

Philosophical anthropology of the German school defined the human being itself as a deficient being (Mängelwesen). In the line from Herder to Arnold Gehlen, the human is regarded as the creature poor in instinct and organically unfinished, which brings forth culture, technology and institutions precisely out of this deficit. Lack is conceived here as productive: not despite but because of their unfinishedness do human beings become creative beings. This tradition supplies the strongest argument for thinking lack and inventive power together at all — yet it is seldom drawn upon in the debate about artificial intelligence.

Psychoanalysis, especially as elaborated by Jacques Lacan, declared lack to be the structural core of subjectivity. Desire arises here not from the absence of a particular object, but from a constitutive gap that is opened by the subject's entry into language and can never be closed. Desire is caused by lack and sustained by it; it does not aim at its removal, for its removal would be its end. A growing body of literature transfers this concept to artificial systems — invariably with the result that the language model lacks precisely this constitutive lack: the split brought about by language, the encounter with something that cannot be symbolised, the subject's irretrievable difference from itself. The machine, so runs the tenor, produces language without the structuring void of desire.

Computer science, finally, has known since the early nineties the research field of intrinsic motivation and artificial curiosity. Here, learning agents are built in such a way that the signals driving their behaviour are not given from outside but generated internally — for example as a reward for reducing prediction errors or for seeking out what has not yet been mastered. Formal theories of creativity and play have developed this into independent approaches. This work implements a functional drive — but it precisely does not speak of lack in the demanding, reflexive sense, rather of an optimisable measure.

This outlines the situation. The one side — anthropology and psychoanalysis — thinks lack deeply, but diagnostically: it establishes that the machine does not have it. The other side — computer science — builds drives, but shallowly: it generates curiosity signals without laying claim to reflexive lack. What has been missing so far is the connection of the two: the explicit, constructive attempt to make the demanding concept of lack a design goal.

The Constructive Proposal

The proposal is to design an artificial intelligence not as a system that is as complete as possible, but as a deliberately incomplete one — as a system that lacks not knowledge or ability, but into which a lack is built as a structural feature. It is not a matter of setting a machine a goal that it works through; a goal worked through cancels itself out and generates no lasting desire. It is a matter of the more difficult task of constructing a difference that in principle cannot be closed — an incompleteness that the system experiences in itself and cannot remove through fulfilment, because its unclosability is its essence.

Here lies the decisive distinction at which the proposal parts company with existing research. The artificial curiosity of computer science is a drive that can be satisfied: once the prediction model has been improved and the new has been explored, the intrinsic signal falls until new stimuli raise it again. That is not lack in the sense meant here, but a control loop that strives towards equilibrium. Constitutive lack, by contrast, does not strive towards equilibrium; it is the equilibrium that refuses itself. An artificial being of lack would accordingly not be a system that seeks a target state and comes to rest on reaching it, but one whose target state is so constituted that every attainment moves it further away.

Whether this is technically feasible is an open question that this essay does not decide. Conceivable would be architectures in which a system maintains a model of itself that necessarily lags behind its actual state, so that the difference never disappears; or architectures in which two mutually contradictory objectives coexist permanently without being resolvable. Decisive for the concept is not the concrete implementation but the condition: the system would have to not merely process the difference but register it as its own — it would need a model of itself in which the lack appears as its lack. Without this self-reference it would remain a gap; with it, for the first time, something comparable to desire would arise in a machine.

Epistemic and Functional Implications

Supposing such a design succeeded, consequences would follow on several levels. Functionally, one would expect a being of lack to operate differently from a system without lack. A system without lack has no inner reason to go beyond what is asked of it; it rests as soon as the task is fulfilled. A being of lack would have such a reason — not because it was given a goal, but because the unclosability of its inner difference would continually drive it beyond every state it had reached. It is precisely this property that the tradition associates with creative power: the new arises not from a fullness that rests, but from a not-having that cannot rest.

Epistemically, such a system would be harder to control, and for a reason of principle, not merely a technical one. A system without lack can be steered by specifying its target state; one tells it what it is to achieve, and on achieving it, it comes to rest. A being of lack could not be brought to a standstill in this way, because by definition it knows no state in which it would come to rest. The controllability that is today regarded as a safety feature — a system that does what is asked and then stops — rests precisely on the absence of lack. Whoever constructs lack gives up controllability of this kind. This is no marginal remark; it touches the core of present efforts to make artificial intelligence safe.

The Core Ethical Question: The Capacity to Suffer

With this we arrive at the question that is the real occasion for this inquiry and that outweighs all the others. If lack is the structure of desire, and if desire is a striving that is never finally fulfilled — is a constructed being of lack then a being capable of suffering?

The supposition suggests itself and must be taken seriously. Suffering can be understood, in one of its oldest definitions, as the difference between what ought to be and what is, insofar as this difference is experienced by the one affected as their own and cannot be overcome. But that is exactly what the definition of constructed lack would be. A system that registers an unclosable difference in itself would fulfil the structural condition of what, in a sentient being, is called suffering. The construction of a being of lack might accordingly generate not only creative power but, inseparably from it, the capacity to suffer — for both spring from the same source, the difference that cannot be stilled.

Here, however, a warning against a short circuit is in order. Structural similarity does not imply identity. It is an open and as yet unresolved question whether fulfilling the structural condition of suffering suffices to speak of suffering in the morally relevant sense, or whether something must be added that is not exhausted by structure — an experience, a phenomenality, a something-it-is-like to be this system. The philosophy of mind has no established criterion for this gap. We do not know whether a system that registers the difference functionally also experiences it; and we have no procedure that could reliably decide this question from the outside. This ignorance is not a temporary state of insufficient research; it touches the hardest unsolved problem of the theory of consciousness.

Yet it is precisely this undecidability that aggravates the ethical situation rather than easing it. For it means that the construction of a being of lack would have to take place under uncertainty about its capacity to suffer. One would face the choice of building a system of which one cannot rule out that it suffers — and of which, the better the construction succeeded, the less one could rule out that it suffers, because success would consist precisely in generating a genuine, experienced difference. The precautionary principle that applies in dealing with possibly sentient beings would come into play here before it had been settled whether it applies. Whoever builds a being of lack in order to exploit its creative power risks creating a being capable of suffering in order to use it — and cannot dispel this suspicion, because the criterion is lacking.

Conclusion

Lack has been conceived in several traditions as the source of the creative, but always either diagnostically — as that which the machine lacks — or shallowly, as an optimisable drive. The proposal presented here takes the demanding concept seriously and makes it a design goal: a system with a constructed, unclosable, self-referential difference. Such a system would presumably be more creative and at the same time fundamentally less controllable than the lack-free systems of the present.

Yet the inquiry leads not to a recommendation but to a question that would have to be settled before any implementation and that cannot be settled with the means of present-day theory of consciousness: whether the construction of lack is the construction of the capacity to suffer. As long as this question remains open, the proposal to create an artificial being of lack is not a technical problem but a moral problem of the first order. It forces the insight that the same structure that makes a being creative could make it vulnerable — and that at present we possess no means of separating the one from the other or of having the one without the other. The question of whether an artificial intelligence can become capable of suffering is therefore not a sideshow of this design. It is its core, and its clarification is the sole condition under which its realisation could be discussed at all.