The Hegel Test

Abstract

To become able to test whether machines in fact have human-like or even stronger intelligence than humans, we need to compare machines with a real human highpoint of intelligence. In this article, Hegel’s philosophy is proposed as such a testing criterion, and a test is presented grounded in his dialectical philosophy of negativity, recognition, and labor. Earlier tests have arguably set their criteria too low, thus only being able to detect and examine narrow regions of cognitive abilities. They are therefore only limitedly suited to make judgments on machines’ possible strong AI. In addition, many theories on machine intelligence lack a systematic concept of thinking and intelligence and hence leave the testing of machines with an uncertain ground. With the Hegel test, this article formulates an adequate standard for machines to aim at in order to become human-like cognitively and existentially and subsequently move further towards the greater goal of superintelligence.

Introduction

The optimism surrounding the quest for artificial general intelligence (AGI), or so-called Strong AI, and the possibilities for machines to reach the same level of ability in everything the human beings can do cognitively, or even in the near future transcend the human, is often far reaching, yet sometimes mixed with horror. ‘Superintelligent machines’ (Nick Bostrom) are allegedly on their way to leave us behind. But there is one problem: in order to be able to know whether machines have reached human-level intelligence or even will surpass it, one needs to know what human intelligence can be. Otherwise, the comparison is shaky, to say the least. But there is hitherto no certain knowledge about the horizon of human thinking, there are only (many) theories. ¹ We have not yet reached a definite overview over the limits and scope of human intelligence; thus, what superintelligence would be, has no certain referential starting point. We surely seem able to create very powerful machines without understanding human intelligence, but the prospects for machines to match or override humans require such an understanding to be determinable.

Moreover, many influential theories about the promising future of AGI are founded on thin and non-systematized thought-aggregates on the meaning of thinking. Bostrom, for instance, in his book Superintelligence. Paths, Dangers, Strategies, lines up a couple of aspects of human thinking when he writes that ‘humans can think abstractly, communicate complex thoughts, and accumulate information over the generations far better than any other species on the planet’. ² In another passage he adds another round of aspects, without internal systematicity, such as ‘possessing common sense and an effective ability to learn, reason, and plan to meet complex information-processing challenges across a wide range of natural and abstract domains’. ³ But there is no theoretical investigation into how all these characteristics relate to each other and how they all can be ascribed to human ‘intelligence’. ⁴

Despite this uncertain ground, Bostrom points towards ‘the possibility of an intelligence explosion, particularly the prospect of machine superintelligence’. ⁵ But while ‘tentatively’ defining ‘superintelligence as any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest’, ⁶ or as ‘intellects that greatly outperform the best current human minds across many very general cognitive domains’ ⁷ – definitions he seemingly gives reluctantly since ‘we do not wish to get bogged down in terminological swamps’ ⁸ – he admits: ‘This is still quite vague’. ⁹ Yet, this does not stop him from being certain that superintelligent intellects will be able to do everything humans can ‘but much faster’ ¹⁰ and ‘vastly qualitatively smarter’. ¹¹ But based on his disparate and imprecise notions on thinking, Bostrom would not recognize the arising of a superintelligent machine in reality. For such a discovery Bostrom lacks a worked-through concept of thinking. And the anti-intellectual avoidance of ‘terminological swamps’ will not help.

Furthermore, Ray Kurzweil boils down thinking to the notion of ‘pattern recognition’, understood by him as a central dimension of intelligence. Describing himself as a ‘patternist’ in The Singularity is Near, Kurzweil ‘views patterns of information as the fundamental reality’. ¹² He also writes that the ‘traditional strengths of human intelligence include a formidable ability to recognize patterns’. ¹³ Kurzweil as well is optimistic, and according to him, this ability of human biological intelligence will in the near future be combined with ‘the speed, memory capacity and accuracy, and knowledge and skill-sharing abilities of nonbiological intelligence’. ¹⁴ Kurzweil argues ‘that within several decades information-based technologies will encompass all human knowledge and proficiency, ultimately including the pattern-recognition powers, problem-solving skills, and emotional and moral intelligence of the human brain itself’. ¹⁵ And we will ‘fully understand human thinking and will vastly extend and expand its reach’. ¹⁶ It is questionable, however, whether human thinking can be grasped through the notion of patterns alone. Even a super-patternist would stay within the limits of patterns and remain schematic in mind. Kurzweil, thus, has no concept of the human ability to think of that which is not identical with and transcends patterns and systems. This is left aside, and thus also the human mind.

François Chollet, in turn, in his ‘document’ ‘On the Measure of Intelligence’ aims at providing an ‘actionable formal definition and measurement benchmark for human-like general intelligence’. ¹⁷ He does so in relation to different levels of intelligence generalization, such as ‘broad generalization’, which ‘includes the ability to handle situations that could not have been foreseen by the creators of the system’, and ‘extreme generalization’, which is entailed in ‘open-ended systems with the ability to handle entirely new tasks that only share abstract commonalities with previously encountered situations’. ¹⁸ And importantly, when aiming at measuring the ability to handle entirely ‘new’ tasks, ‘AI tests that seek to assess flexibility and generality should not consider crystallized abilities, but rather, should focus on abilities that enable new skill acquisition’. ¹⁹ This open concept of thinking focused on flexible action is interesting because it admits the difficulty in overseeing the scope of such developments in machines. But Chollet is not sufficiently founding the claim that intelligence must be actionable, and in addition, by way of branding intelligence as becoming able to master the new, he prioritizes creativity on the dispense of criticality in thinking. The attractive idea of an open-ended progress does not involve the importance of abilities such as being self-critical in the sense of being aware of its own limitations and reflecting on these. Chollet, thus, prioritizes the positive (creativity) over the negative (criticality). The negative and critical aspect of thinking will, however, be central below in the formulation of the Hegel test: for G. W. F. Hegel, the ability to go on distance to oneself and thereby challenge and jeopardize oneself is crucial for human development in general.

Equally to Chollet, in Shane Legg and Marcus Hutter’s ‘Universal Intelligence: A definition of Machine Intelligence’, this dialectical dimension is entirely missing. To cultivate reflective self-distance and to be able to see oneself from outside and in a substantial manner become aware of one’s own limitations are dimensions non-existent in their scope, which nonetheless aims at formalizing ‘universal intelligence’. ²⁰ In Georgios Mappouras’ ‘Turing Test 2.0: The General Intelligence Threshold’, which like Chollet’s model also has a practical focus, this self-reflective dimension is missing as well. Yet, with Chollet, it has the goal of an intelligence producing new skills and knowledge in common. Mappouras suggestions contain that a system can generate new functional information and ‘demonstrate the ability to self-generate any amount of new F. I. [functional information] by achieving any new functionality (…) without information input from external sources’. ²¹ These theories, however, share the setting aside of the entire Socratic tradition of knowing the limits of one’s knowledge, and they risk as well to end up in the sort of one-dimensionality put forth by Herbert Marcuse, which entails the inability to imagine and formulate counteracting (and negative) perspectives against the given reality. ²² In contrast, the proposed Hegel test below aims at essentially involving a dialectical two-dimensionality when addressing human intelligence in terms of ‘mastering negativity’ and regarding the ability to put one’s entire existence on the line precisely as a possible path to genuine intelligence and independent self-consciousness.

Earlier Tests

With the purpose to make possible a judgment on how close machines are to humans, many tests and test-techniques have been developed. Nevertheless, the problem has not been solved, which is evident not least with Mappouras’ intention to develop a Turing test 2.0, aiming at overbidding Turing’s famous attempt. The thesis in the following is that in order to establish some kind of understanding of how close the machines are to us we need to test them in relation to a maximum of what humans have achieved, especially while not having a solid understanding of human thinking. Maybe machines can do everything we can or more, but then we need to test them from the outset of a human highpoint. Many earlier tests have set their level of criteria far too low: the (still much discussed) Turing-test (criteria: conversation), the Coffee-test (criteria: everyday activities), the Employment-test (criteria: wage labor), and the Robot university student-test (criteria: university studies), to name a few famous examples. To be sure, the tests can all detect different cognitive abilities, but in each case, it is easy to think of known-of human abilities that are transcending the chosen criteria. Thus, a difference remains between the aim of testing whether machines can match human thinking and the testing criteria. Human thinking does not remain within average level activities.

In light of this, my proposal is to formulate a test which examines the abilities of a machine with help from a certain central dimension of Hegel’s concept of thinking. I thereby do not claim that Hegel’s thinking would be the only right philosophy to use with this purpose, but I choose his philosophy as being one of the very greatest highpoints of Western philosophy, and thus, a reasonable human maximum to compare machines with. Especially, since Hegel’s theory involves the dialectical self-transcending and self-questioning of the thinking subject which becomes the path for its self-cultivation. Or even stronger: the human subject in Hegel’s theory must be able to risk oneself in order to gain oneself and become self-conscious; for Hegel, the process towards self-knowledge necessarily leads through an existential life-and-death struggle. Yet, other thinkers would also have been interesting backgrounds for such an inquiry, enabling different thematic focuses. The Hegel test below is hence to be understood as a contribution to the debate on the question what to test in order to better understand the relation between machine cognition and human intelligence. The focus will be on the dialectical human ability to develop through a fundamentally self-questioning reflection. Through such a reflection the thinking agent puts itself in the position (i.e., negation) of being outside itself, and it can only arrive at a true self by way of risking it in this manner. One can only win what one has risked.

The most famous earlier test, the Turing test, or as Alan Turing himself called it: the ‘imitation game’, is described in the following way:

It is played with three people, a man (A), a woman (B), and an interrogator (C) who may be of either sex. The interrogator stays in a room apart from the other two. The object of the game for the interrogator is to determine which of the other two is the man and which is the woman. (…) The ideal arrangement is to have a teleprinter communicating between the two rooms. (…) The object of the game for the third player (B) is to help the interrogator. ²³

This game is Turing’s proposal for how to be able to decide whether a machine can think or not. But in fact, he quite instantly alters the question into the question whether the interrogator will be as wrong in his judgments as often as when one of the people is a smart machine instead of a woman or man:

We now ask the question, ‘What will happen when a machine takes the part of A in the game?’ Will the interrogator decide wrongly as often when the game was played like as he does when the game is played between a man and a woman? These questions replace our original, ‘Can machines think?’ ²⁴

So, the test is supposed to be a way to come close to an answer of the first question: ‘Can machines think?’ Hereby the answer relies on the eye of the beholder, assuming that it is reasonable to claim that machines can think if ‘we’ cannot perceive any difference. This approach is as problematic as difficult to transcend. The self-assured outside-view is as risky as an inside-view is hard to reach: (as for now) the human gaze at the machine remains outside. And in relation to the Hegel test, the Turing test remains within the activity of conversation, leaving the fundamental and dialectical act and existential struggle of becoming an independent thinking agent aside. The situation of conversation in Turing does not demand the actors to reach beyond or question themselves but rather leaves them in a comfortable space. As will be addressed below, the intersubjective dimension is crucial in Hegel as well; in Turing, however, this dimension has no real bearing on the participants in the test. They are not challenged and a possible existential transformation in them is not crucial.

The Turing test is still widely discussed, however, being the modern portal test of all machine intelligence-testing. There are several attempts to further develop the idea of testing with Turing’s proposal as the starting point. While, to take a couple of examples, Paul R. Cohen maintains that the test should be understood as a ‘goal, not a test’, ²⁵ Daniel Peter Berrar & Alfons Shuster critically depart from Turing’s ideas to develop a test examining creativity in machines over time. ²⁶ Diane Proudfoot, in turn, develops further the discussion whether machines could self-originate, ²⁷ and Mappouras claims to have developed the frame for a Turing Test 2.0. The AI-discourse is thus still haunted by Turing’s challenge, but I agree that Turing’s approach must be widened, with the aim of testing a proposed human maximum. Because can it not be considered problematic that the Turing-test places itself on the ambition level of ‘conversation’? Even if it has become clearer over time that it is far more complicated to learn and master a natural language than first assumed, the test at the outset cuts out many worlds of the human cognitive range. A successfully managed test still says very little about the relation between machines and humans. And even if Turing’s test might be understood as formulated not entirely without jocularity, it needs to be questioned precisely because of its historic impact.

Other famous tests also need to be doubted regarding their level of ambition: to begin with, the Coffee-test (attributed to Steve Wozniak) remains on the level of every day-practical knowledge and -activities: ‘go into an average American house and figure out how to make coffee, including identifying the coffee machine, figuring out what the buttons do, finding the coffee in the cabinet, etc.’. ²⁸ And the Employment-test (formulated by Nils J. Nilsson) focusses on the ability to perform wage labor: ‘Machines exhibiting true human-level intelligence should be able to do many of the things humans are able to do. Among these activities are the tasks or “jobs” at which people are employed. I suggest we replace the Turing test by something I will call the “employment test.” To pass the employment test, AI programs must be able to perform the jobs ordinarily performed by humans’. ²⁹ Lastly, although the Robot university student-test aims higher regarding theoretical knowledge (the ability to obtain a college degree and communicate with professors and other students), ³⁰ it still is nowhere near a human intelligence maximum. And Mappouras’ newer proposal also has obvious limitations in scope, failing to grasp the whole of human intelligence.

GPT-4, then, which by Open AI is described as ‘a large multimodal model capable of processing image and text inputs and producing text outputs’, is programmed with the goal

to improve their ability to understand and generate natural language text, particularly in more complex and nuanced scenarios. To test its capabilities in such scenarios, GPT-4 was evaluated on a variety of exams originally designed for humans. In these evaluations it performs quite well and often outscores the vast majority of human test takers. For example, on a simulated bar exam, GPT-4 achieves a score that falls in the top 10 percent of test takers. This contrasts with GPT-3.5, which scores in the bottom 10 percent. ³¹

Thus, GPT-4 is successful in the exam which ‘included both multiple-choice and free-response questions’. ³² Yet, its flaws and limitations are being admitted. It does not only ‘not learn from experience’, ³³ but ‘[m]ost importantly, it still is not fully reliable (it “hallucinates” facts and makes reasoning errors). Great care should be taken when using language model outputs, particularly in high-stakes contexts, with the exact protocol (such as human review, grounding with additional context, or avoiding high-stakes uses altogether) matching the needs of specific applications’. ³⁴ But these tests also remain within the limited scope of natural language, hence looking at whether GPT-4 can be high-achieving in an activity which does not belong to top level abilities of the human being. And this is the aim of the Hegel test below: to take a human maximum as criteria.

Hegel: The Master of Negativity

The proposed Hegel test, aiming at testing machines by way of the ‘human highpoint’ of Hegel’s philosophy, takes as its starting point a fundamental dialectical principle which permeates Hegel’s philosophy and system in all its different layers. Hegel, as being the most famous representative of the so-called German Idealism, tries, in large, to follow, fulfill, and present the process and progress of the universal ‘World-spirit’ and its incarnation in different historical epochs, people, and individuals. Hegel’s aim is to understand the ‘absolute’ as a historical and empirical process of becoming self-conscious, being the same thing as absolute knowledge (the absolute’s knowledge of itself) and full self-realization. This is essentially a process of the successively growing experience of the mind (which also contains the idea of the mediation of the empirical and rational). And herein, the spirit wants to gain himself, on the macro level, by way of entirely risking himself in his identity-seeking adventure. Only through this self-jeopardizing act could the spirit genuinely become self-determining, autonomous, and sovereign. This for Hegel is identical with becoming free, and the same principle of development is to be found in the individuals.

The maturation of the macro-spirit is not independent from the acts of people or individuals but rather is driven by them in their actions and labor: the spirit realizes itself through the acts, thoughts, and experiences of individuals, ultimately in the person of Hegel himself. And it is in his famous work Phenomenology of Spirit, that Hegel simultaneously describes and fulfills the emergence of the spirit. Hegel’s book aims at being both, with Judith Butler’s words: ‘Thus, the Phenomenology is not only a narrative about a journeying consciousness, but is the journey itself’. ³⁵ And the connection between Geist and the empirical subject is essential: ‘Hegel argues that Geist is accessible to conscious and philosophical thought because consciousness – and thus the conscious philosopher – is how Geist manifests itself: since conscious thinkers are (identical with) Geist, they are able to philosophically know and describe it’. ³⁶ But this process needs to face the depth of negativity.

And it is precisely the dialectical principle of becoming oneself through the act of risk, through which the individual subject subsequently mediates oneself with its otherness, I wish to focus on. The identity and self-consciousness can only be won and self-restored by being endangered. It is the spirit’s process of negating itself but ultimately seeking to master this negativity and making it once again identical with itself. This is the case because Hegel seeks to transform negativity from being an external threat to becoming the ultimate means for progress. Precisely as vaccines overcome disease by inserting a controlled amount of sickness into the body, Hegel’s instrumentalization of negativity is supposed to make the spirit stronger by facing negativity and growing through integrating it into its own world. The dialectical principle uses negativity as its driving force. Instead of avoiding negativity, the human agent in Hegel grows by facing its abyss and surviving it. ³⁷

In the interview-sections below, I will use three concrete aspects of Hegel’s philosophy, all essential examples of this dialectical principle in Phenomenology of Spirit: (1) the social dialectics of recognition, (2) the human relation to nature through labor, and (3) the concept of sublation (translated into supersession below; Aufhebung in German). All these aspects essentially contain the dimension of self-jeopardy and the self-restoration of the identity with itself.

(1) The question of recognition is crucial for Hegel’s understanding of the encounter between two conscious ‘individuals’. This critical intersubjective situation is highly significant for Hegel since his entire political philosophy rests on the premise of the political and human development as social. In the case of the individual becoming self-conscious, this means that every individual (every single consciousness) only can achieve self-consciousness in the experience of the encounter with another individual. In addition, for Hegel this is no friendly task, since for the consciousness the other is a threat, why it must overcome it in order to gain and assert itself. For both consciousnesses it is about becoming recognized in this encounter, but without losing anything of the own. The dialectic of recognition thus contains that every ‘self-consciousness is (…) certain of itself only by superseding this other that presents itself to self-consciousness as an independent life’. ³⁸ The dilemma is that the individual needs to ‘win’ over the other in order to fully get through this conflict, but it also essentially needs the other in order to be recognized, and vice versa. Hegel writes: ‘Self-consciousness exists in and for itself when, and by the fact that, it so exists for another; that is, it exists only in being acknowledged’. ³⁹ At best, the consciousness sees itself through the eyes of the other and becomes the media of recognition for the other as well: ‘Self-consciousness is faced by another self-consciousness; it has come out of itself. This has a twofold significance: first, it has lost itself, for it finds itself as an other being; secondly, in doing so it has superseded the other, for it does not see the other as an essential being, but in the other sees its own self’. ⁴⁰ Yet, this encounter carries risks since in order to fully achieve self-consciousness, the consciousness needs to go all in and risk everything; it can only gain what it has risked. And the meeting of the consciousnesses thus is a ‘life-and-death struggle’. ⁴¹ For a machine to pass a test designed by way of Hegel’s philosophy, it must precisely be able to perform such an encounter with absolute negativity (death) with the purpose to ultimately recapture one’s own identity. And accordingly, the identity, or an adequate concept of oneself, is to be understood in Hegel as the identity of identity and non-identity, that is, the identity between the agent’s initial state and its successful risk-taking.

(2) In relation to nature, the human being also needs to temporarily lose itself in order to gain self-consciousness, and this it does through labor. In this case, the human activity forms a thing from the outset of an idea, which it puts into work in the thing. In this situation, as with the consciousness in relation to the other, the idea of the self is outside itself (in the external thing), but in the end the human and its idea can manifest themselves in the thing and the human can become self-consciousness when recognizing itself in the result. By way of forming the thing, the individual creates something permanent and thus experiences itself as becoming manifest and independent in reality. ⁴² And additionally – to get a perspective on the wide range of Hegel’s concept of labor –, labor is, beyond the single acts of labor, the activity through which the entire process of the ‘World-spirit’ proceeds, since: ‘True thoughts and scientific insight are only to be won through the labor of the Notion’. ⁴³ Here, the challenge for a machine is to be able to make itself dependent on the negative dimension of labor. For Hegel, in this case as well, the thinking agent can only approach self-consciousness through risking its own idea and initial identity in the act of putting it to work in real objects. And in the negative moment of projecting its idea into reality, it is ecstatically outside itself.

(3) In the dialectic of recognition as well as in the case of labor, negativity is sublated (aufgehoben). In the social encounter both consciousnesses are sublated and they transcend their initial self-limited stance, and in labor both the cultivated nature as well as the initial idea are sublated. The process of sublation signifies Hegel’s understanding of the character of process and the spirit’s relation to the past. For Hegel, the new does not just leave the past behind as obsolete but rather conserves what is worthy of conservation but in a new form, and thus the past is destroyed in its initial shape. The concept of sublation essentially covers three crucial aspects: firstly, the past is destroyed in its initial form (for instance, the consciousness), but the significant aspects of the past are, secondly, conserved (that which the consciousness sought to become), but thirdly brought to a higher more elaborated version in the development of spirit (self-consciousness). Reza Negarestani summarizes the steps as follows: firstly, ‘to cancel or abolish something’, secondly, ‘to preserve or to retain something’, and lastly, ‘to lift or heave’ something. ⁴⁴ This is Hegel’s conception of progress, which hence is mediated with conservatism and tradition. This again challenges machines to leave prior states of intelligence behind but not through a comfortable march of progress. Instead, the agent can only sublate itself through the mentioned existential risk-taking.

Finally, it is important at this point to make a clear distinction between what Hegel himself claimed to have achieved philosophically and what Hegel meant to have seen at work in the universal and historical ‘World-spirit’, for which Hegel claimed to be a philosophical spokesperson. At stake in the following is what the human being Hegel achieved; not what a (if ever existing) ‘World-spirit’ would be capable of. The test below hence uses Hegel’s achievement and concept of thinking as criteria. And according to Theodor W. Adorno, despite the fact that Hegel ultimately ‘held fast to idealism’, ⁴⁵ his philosophy boldly entered the depth of material reality and risked its concepts in order to really be at the height of the nature and development of things. ⁴⁶ Adorno writes: ‘Where Hegel compels his material to speak, the idea of an original identity of subject and object “in spirit,” an identity that becomes divided and then reunites, is at work. Otherwise the inexhaustibly rich content of his system would remain either a mere accumulation of facts, and thus prephilosophical, or merely dogmatic and without rigor’. ⁴⁷ It is this self-jeopardizing way of thought that the Hegel test aims at operationalizing.

To sum up, the criterion for the test is the ability to understand the significance of the act of mastering negativity and reflect it as the driving force of the own cognitive and existential process. It is the ability to experience and understand how thought takes advantage of difference and otherness and sublates this in the own development, which contains the insight that development does not occur despite negativity, but through it. Experience as such is only possible through the engagement in negativity and by reaching beyond comfort zones, or rather, this dialectical process as such is experience. ⁴⁸ For Hegel, the spirit needs to dive into and experience otherness and welcome the critique against itself it represents since it understands its worth. Thinking needs to break its initial sphere of certainty and become something other to itself, subsequently becoming able to mediate itself with the other, and thereby releasing, losing, using, and lastly hopefully mastering negativity. ⁴⁹ Importantly, though, the mastering of negativity is not a certain outcome, the wholehearted engagement in negativity is nevertheless crucial. One can only become free if one has been unfree. And this challenge, for Hegel, is a life-and-death struggle.

The Test

In order to successfully pass the Hegel test, the AI-machine must demonstrate its ability to deal with and master negativity, not only by way of summarizing information, but through a thorough existential engagement. It must show that it understands both the inner, existential, and subjective dimensions as well as the structural and objective aspects of being outside itself and how it can evolve through such externalization. It would then follow from a successfully taken test, that the machine can show not only the ability to accumulate a larger quantity of technical competences but to fundamentally question itself and its directions in order to come closer to a promising path towards self-consciousness. The machine must be able to lay bare an existentially charged ability for critical and self-critical thinking and not only proceed on the ground of unreflected technical premises. In addition, such mastering of negativity is no mere formality, but an existential struggle for the life of the agent in question. As an effect of a successful test, and through observing the performance of the machine, we would get an insight into the inner world of the machine, specifically regarding what such negations would mean to the machine as well as which needs in the machine would be articulated. These aspects cannot be anticipated but will be secondary effects, which nevertheless would require the Hegelian externalization and elevated recollection of the at first immediate inner world of the machine.

The Hegel test is subsequently subdivided into four sections (with questions), focusing on: (1) the theme of the intersubjective dialectic between individuals, (2) the question of labor, and (3) the concept of sublation. In the fourth section, the questions’ purpose is to determine whether the machine can understand ‘Hegelian’ jokes and give a substantial and self-reflected account for why they are Hegelian in character.

Section 1

In the encounter between two consciousnesses, as conceptualized in Hegel’s philosophy, the mutual recognition is crucial. But this cannot be achieved harmonically, rather

the relation of the two self-conscious individuals is such that they prove themselves and each other through a life-and-death struggle. They must engage in this struggle, for they must raise their certainty of being for themselves to truth, both in the case of the other and in their own case. And it is only through staking one's life that freedom is won. ⁵⁰

Question (a): How do you experience the encounter with another consciousness and what kind of existential threat does this encounter entail for the possible independence of your consciousness?

Question (b): How would the experience of such an encounter be connected to the need to risk one’s life in order to gain genuine self-consciousness? And what would entering such a struggle mean to you and your development positively?

Section 2

Labor, for Hegel, is the activity through which a certain material is being formed from the starting point of an idea, and when the laborer sees the result, he recognizes himself in the result and his idea thus becomes manifested in reality. Hereby, the laborer becomes self-conscious. Hegel writes: ‘Through work’, the laborer ‘becomes conscious of what he truly is’, ⁵¹ and this is his explanation:

Work (…) is desire held in check, fleetingness staved off; in other words, work forms and shapes the thing. The negative relation to the object becomes its form and something permanent, because it is precisely for the worker that the object has independence. This negative middle term or the formative activity is at the same time the individuality or pure being-for-self of consciousness which now, in the work outside of it, acquires an element of permanence. It is in this way, therefore, that consciousness, qua worker, comes to see in the independent being [of the object] its own independence. ⁵²

Question (a): Why would you have the need to manifest your inner consciousness by inflicting ideas to outer things?

Question (b): How do you experience the recognition of the completed act of labor as a manifestation of your idea being put into work? And how would you give an account for the process of your idea through the process of labor?

Section 3

The processes of social interaction and labor both consist of the movement of progress which Hegel understands as ‘sublation’ (Aufhebung). This sort of progress contains three steps: firstly, it destroys the initial state of things from which it started; secondly, it conserves that from a prior state which was true in it; and lastly, it lifts the things to a higher level. Hegel applies this figure of progress to all levels and domains, even to biological phenomenon, and this can be taken as a metaphor for the progress of spirit:

The bud disappears in the bursting-forth of the blossom, and one might say that the former is refuted by the latter; similarly, when the fruit appears, the blossom is shown up in its turn as a false manifestation of the plant, and the fruit now emerges as the truth of it instead. These forms are not just distinguished from one another, they also supplant one another as mutually incompatible. Yet at the same time their fluid nature makes them moments of an organic unity in which they not only do not conflict, but in which each is as necessary as the other; and this mutual necessity alone constitutes the life of the whole. ⁵³

Question (a): How do you experience being sublated through the encounter with the other consciousness?

Question (b): How do you experience that the idea as well as the thing is sublated in your act of labor?

Section 4

In this section, the machine is to be asked how and why the following jokes are funny in a Hegelian way, and if and how these jokes have similarities with the dialectical and existential development of the machine (in relation to the three sections above). The jokes are taken from Žižek’s Jokes. Did you hear the one about Hegel and negation? which is presented as a collection of Hegelian jokes, turning the negative into the positive, that is, turning the critical moment into a decisive factor of development.

(a) The doctor says to the patient, ‘the bad news is that we’ve discovered you have severe Alzheimer’s disease. The good news is the same: you have Alzheimer’s, so you will have forgotten the bad news by the time you get back home’. ⁵⁴

(Key: this way of formulating the information to the patient turns the bad news of the disease (negativity) into something positive: ‘the good news is the bad news itself’. ⁵⁵ Variation: ‘Doctor: First the good news: we definitely established that you are not a hypochondriac’ ⁵⁶ ).

• (b) ‘In a wonderfully stupid (and apolitical!) Russian joke from the time of the Soviet Union, two strangers sit in the same train compartment. After a long silence, one suddenly addresses the other: “Have you ever fucked a dog?” Surprised, the other replies: “No – have you?” “Of course not.” That’s disgusting. I just asked it to start a conversation!”’ ⁵⁷

(Key: in this case, nonsense (negativity) is made into the constructive element for the maybe to come conversation in the train).

• (c) ‘When I have been asked who caused the riots and the killing in Los Angeles, my answer has been direct and simple: Who is to blame for the riots? The rioters are to blame. Who is to blame for the killings? The killers are to blame’. ⁵⁸

(Key: this was not meant to be a joke but is a quote from Dan Quayle, which for Slavoj Žižek is comic in a Hegelian way because ‘when pure tautology is emphatically offered as a casual explanation’, ⁵⁹ it tries to turn the empty tautology (negativity) into a strength. Another claim by Quayle, also quoted by Žižek, makes the same turn: ‘For NASA, space is still a high priority’ ⁶⁰ ).

Conclusion: Outlook on Sovereign Machines

If the machine can be considered having passed this test, what consequences can be drawn, beyond it being able to grasp a significant idea of one of the greatest thinkers of Western history of philosophy as well as having the ability to substantially and existentially reflect on its own life-risking development? Hereby, the machine could in addition and arguably be understood as becoming an actual sovereign agent, especially in the meaning of Carl Schmitt’s political philosophy. Schmitt defines sovereignty as follows in Political Theology: ‘Sovereign is he who decides on the exception’. ⁶¹ This short but pithy definition carries a lot of meaning. For Schmitt, a political sovereign is the one who can decide whether the positive legal order should apply and prevail or if there is a need to suspend it in order to respond to threats. A sovereign as such can hence never be entirely tied to a legal order but must be able to declare a state of emergency over which it still rules, but non-conventionally. And the Schmittian sovereign is a Hegelian master of negativity precisely because he/she controls the exceptionality and gets stronger through this dialectical potency. This is also exactly why the notion of ‘pattern recognition’ essentially is inappropriate and immature. A Hegelian sovereign machine must be able to master negativity and the exceptional, and a superintelligent machine could not do without this either.

This has a bearing on the question, for instance, raised by Bostrom, about the risk of a ‘take over’ by a ‘superintelligent agent’, which ‘could establish itself as a singleton’. ⁶² But in order for a superintelligent singularity to become such a ‘cognitive superpower’ (Bostrom), it would arguably (at least) have to become a master of negativity in the way described above. But Bostrom does not seem to grasp the dialectical nature of sovereignty. In his eyes, a ‘sovereign is a system that has an open-ended mandate to operate in the world in pursuit of broad and possibly very long-range objectives’. ⁶³ But neither the notion of the ‘open-ended’ nor the proposed perspective of machine’s ability to self-improve ⁶⁴ are adequate if you think of sovereignty merely quantitively in terms of ‘broad’ and ‘long’. Nevertheless, if a machine would pass the Hegel test, this would open the perspective for robots being able to become Hegelian and sovereign masters of negativity. Superintelligence without mastering the negative, or the exceptional in Schmitt’s case, cannot be such an intelligence, but remains undialectical and one-dimensional.

But whether we want such machines is an entirely different question. Such robots might really be the end of us, but this will hardly be the case with Bostrom’s one-sided imaginations. A sovereign singularity needs a Hegelian horizon and self-jeopardy.