The relationship between the subject and an increasing variety of artificially intelligent (AI) chatbots has proven a unique point of contention within psychoanalytic debates (Everitt, 2023; Johanssen, 2023; Murphy, 2023; Rouselle, 2023; Žižek, 2023a). In fact, outside of psychoanalysis, concerns regarding the technology have been shared by those tasked with its development. In May 2023, Geoffrey Hinton, a British-Canadian cognitive psychologist and computer scientist, left his role at Google, citing fears regarding the dangers of AI technology on society and humanity. For Hinton, the potential for the technology to be misused, as well as the wider impact of its adoption on jobs and employment, presented worrying concerns regarding the informational capacity of AI chatbots, including their ability to share and accumulate knowledge on a scale far beyond the human individual (Kleinman & Vallance, 2023). The extent to which this knowledge may be manipulated by AI technologies, without human instruction, bears witness to the suggestion that AI technology could eventually become more intelligent than its human developers.

Certainly, the concern that AI will outstrip the intelligence of its human creators is itself a useful hubristic fantasy, which implicitly assumes that we are capable of inventing such a ‘complete’ AI, either now or at some point ‘in the future’. Without dismissing the level of human creativity and ingenuity that has gone into developing the technology, the possibility of a rogue AI goes someway to ignoring the many cases where computer technologies have failed due to some programmable human error.

More to the point, when AI technology remains a human invention, research into the desire of its developers and users should be a significant point of inquiry (Millar, 2021). Alongside those surveys that have sought to critique assumptions regarding whether AI technologies are becoming ‘human’ (Johanssen, 2023), I argue that, for now, our relations with the AI chatbot rest not so much on its function as a source of knowledge, but on the desire for the subject not to know.

Accordingly, in the following account, I briefly trace both the psychotic and perverse positions to help outline the psychosocial structures that the AI chatbot presents. By reflecting critically on how such technology impacts upon the subject’s ethical responsibility, I couch this discussion in a consideration of the extent to which the AI chatbot serves to expose the relation between two forms of subjectivity: the subject of knowledge and the subject of desire.

Paranoia, Perversion and the AI Chatbot

One major software development has been to enhance the AI chatbot’s capacity to mimic human conversation. Here, one can easily begin a conversation with an online chatbot, from which the replies that one receives closely echo the responses of a fellow human being. It is for this reason that debates on whether the chatbot has an unconscious or if it in fact represents a new externalized unconscious have culminated around the apparent ‘threat to the social bond’ that the AI chatbot reveals (Murphy, 2023). This degradation of the social bond, or symbolic order, further entrenches our turn towards the imaginary and the apparent lack of prohibition. In our use of the AI chatbot we can, potentially, discuss any topic, receive any answer, and exploit whatever knowledge that the chatbot presents. What is lost in this failure of prohibition, however, is castration – the foreclosure of which Rouselle (2023) relates to the loss of the unconscious space prescribed to dreaming: today, our dreams are externalized and brought to bear online; shared and manipulated in the digital space where AI resides.

With repression stalled and castration lost, we are left with either a path of perversion, and the disavowal it performs, or a psychosis premised on the foreclosure of the Name-of-the-Father, the very source of authority. Indeed, where ‘paranoia and perversion dance in harmony’ (Rambatan & Johanssen, 2021, p. 112),Footnote 1 it is in the seduction or incomprehensibility of the Law that the persecution of the psychotic or the creation of the pervert is established (Marini, 1986/1992, p.77). In fact, whether our relations with the AI chatbot denote a perverse or psychotic position, neither presents a negation of the Other. While the pervert acknowledges their online manipulation by the Other (despite its almost-human replies the chatbot is nevertheless just an artificial application predicated on a finely tuned algorithm), for the psychotic, it is the paranoia that there may be a nefarious, all-knowing Other, working to exploit our chatbot interactions, which proves distressing. It is this knowledge that remains routed through the Other, enveloped in a paranoia that seeks the subject’s persecution (Lacan, 1966/2006, pp. 82–101).

It is in opposition to this paranoia that Žižek (2023b) contends that ‘it is perversion (or père-version, “version of the father,” as Lacan puts it) and not psychotic isolation that characterizes the AI’. That is, what is disavowed in our relations with the AI chatbot is the enjoyment that we receive from evoking the stupidity of the machine – its mistakes, inconsistent replies and potential for repetition. In so doing, we disavow the very fact that it is ‘us’, the user, who provokes the chatbot’s replies. Where the psychotic errs is in its positioning of the AI chatbot as the ‘subject supposed to know’: a ‘mistake’ that Žižek locates in the confusion that arises when ‘the virtual/symbolic big Other and the actual digital big Other’ coincide (2020, p. 160, parenthesis removed). Here, ‘a materially existing big Other’, that is, the digital infrastructure comprising the AI chatbot, is mistakenly perceived to encompass some ‘divine Other’ (Žižek, 2020, p. 161).

Žižek’s (2020) critique can be considered for elucidating on how our interactions with the chatbot hold a certain fascination: a desire for truth or knowledge, which, in its position of supposed authority, proves particularly alluring for the subject. As Hook asserts, ‘the goal of supposed knowledge and truth, is a powerful motor of subjectivity, hence Lacan’s reference to the Other … as “the subject supposed to know”’ (2018, p. 32). Located in this position of the Other, does the AI chatbot not occupy such a position of authority? A cache of knowledge, which, beyond our more trivial and playful concerns (asking the chatbot to create a joke, for example, or, for the more perverse, to find ways of encouraging it to expel certain discriminatory remarks), functions as the very space in which we position ourselves in relation to knowledge?

The Subject of Knowledge

The subject of knowledge relies upon the very distinction that it prescribes between itself (the subject) and the world outside (the object(s) it wishes to gain knowledge of). It is for this reason that ‘the subject of knowledge begins in a state of ignorance and approaches the world as an object of inquiry that exists prior to and apart from that inquiry’ (McGowan, 2012, p. 40). While the most common example of such a subject is, as McGowan asserts, the scientist driven by scientific discovery, we nonetheless ‘approach the world as subjects of knowledge insofar as we tend to picture ourselves as simply wanting to know things that appear unrelated to us, even if it is a question of our interest in the lurid behavior of celebrities’ (2012, p. 40). So common is this assumption that it can be said that our own personal development is fundamentally dependent on us knowing more.

Serving as the companion to the subject of knowledge, there is the subject supposed to know, or, as Schneiderman asserts the ‘supposed subject of knowledge’ (quoted in Evans, 2006, p. 199) – in this case, the AI chatbot. This supposition of the AI chatbot underwrites the very way in which the subject of knowledge encounters the utility of the world around them. It is here that the functionality of the AI chatbot is expressed in questions that, for example, may seek to know the best restaurant in a specific location or where best to stay during a trip away (both of which the AI chatbot will helpfully answer).Footnote 2 This is also apparent when we rather lazily ask the AI chatbot to draft a cover letter or an email reply. While the AI’s responses may appear less fathomable than a Google search, in either case, it is the utility of the chatbot that proves undoubtedly fascinating, and, to a certain extent, more direct, or perhaps, more ‘real’ than a simple Google search.

There is, therefore, a certain ‘promise’ that underwrites the chatbot’s utility and the questions we ask: ‘the promise of artificial intelligence’ to helpfully provide the knowledge which we seek and the potential to know (Murphy, 2023). Here Murphy asserts that, ‘from the pandemic onward, we see iterations of these fix-it-all hacks that can give the promise either of comfort now, or in the future. These range from Virtual Reality to Crypto to Artificial Intelligence’ (2023). Accordingly, while ‘the subject of knowledge is a hopeful subject, a subject invested in future possibilities, … it is precisely this investment in the future that is the source of the subject’s dependence’ (McGowan, 2012, p. 65). This investment bespeaks a dependence on the Other that is relayed in future possibilities that hold the promise of reaching truth.

Yet, what proves to be the generative factor in such a search for knowledge is the doubt it relies upon; that is, ‘the procedure of skeptical doubt suspects everything but its own suspicion, and this results in gaps in the knowledge it produces’ (McGowan, 2012, p. 47). These gaps become visible when we consider that what the capacity to doubt reveals is the very failure to doubt one’s very doubt (in sum, the knowledge of one’s doubt cannot be doubted). It is in this failure that a gap within (our) knowledge resides. Indeed, while psychoanalysis is steered towards confronting the gap in one’s knowledge – one’s knowledge of oneself, for example – it is this gap that is either disavowed (perversion) or foreclosed (psychosis) by the subject of knowledge. In fact, what emerges in these psychoanalytic structures is how the threat of knowledge is linked to the subject’s jouissance (Mills, 2019). Although our enjoyment rests upon the very restrictions and limitations we impose, whether this be through actively disavowing the prohibiting authority, or seeking to uncover and know the authority working to maintain these very restrictions, in either case, it is in accordance with the impenetrability of the Other that the unconscious is averred. Although the Other remains an unknowable quality that routinely leaves the subject in the position of interpreting, guessing and seeking to know (Hook, 2018), this is always predicated on the fact that the Other is incomplete and lacking.

Furthermore, it is this lack that underscores ‘the imperfection, or in-built stupidity of any AI system’ (Johanssen, 2023). What remains central here is the ‘in-built stupidity’ that underwrites the chatbot. Accordingly, while ‘OpenAI’s chatbot can generate new ideas, poetry, lines of code, anything that language can express’, as Johanssen explains, this is ‘based on probability models and large training datasets’ (2023). In so doing, the chatbot is coded to express both human and non-human qualities. ‘This’, Johanssen argues, ‘is a deliberate move on the part of the developers and constitutes a defense mechanism which cautions against all too human characteristics of AI when emphasising its non-humanness’ (2023).Footnote 3

It is perhaps difficult to determine whether these imperfections are deliberately built into the algorithm, bearing in mind that the developers would, in such cases, be acknowledging the fact that they are intentionally – and consciously – undermining their own work. More to the point, this also runs counter to the fact that AIs work by ‘learning’ from their very mistakes (Felton, 2023). Accordingly, whether the AI’s mistakes are purposely programmed or merely genuine mistakes that occur as part of the AI’s development – and, here, we must remember that the AI remains a human invention and, thus, open to human mistakes – in either case, this does not dislodge the fact that mistakes are made; that, in short, the AI lacks. To this end, we can consider how it is in view of these mistakes that Johanssen asserts that ‘ChatGPT symbolises the move from (Lacanian) desire to drive’ (2023). It is for this reason that:

The fantasy of the big Other is frequently shattered because ChatGPT makes mistakes, invents things, or freely admits that it lacks knowledge on a particular question. The subject knows that it cannot fulfil a therapeutic or authoritative function but is itself lacking in a way. Unlike the fantasy of its developers, ChatGPT actually reveals the impossibility of desire. (Johanssen, 2023)

Extending this line of inquiry, we can consider how the revelation of this impossibility functions to expose the subject of desire.

The Subject of Desire

For Lacan (1975/1998), psychoanalysis reveals that the subject lacks a genuine desire for knowledge. Although we may take pleasure in learning new things, as well as debating with ourselves and others why certain relationships or personal circumstances may not be satisfactory, when set against the enjoyment that inheres in these symptoms it is the subject who ‘wants to know nothing more about it’ (Lacan, 1975/1999, p. 105). Importantly, this suggests that rather than following a path towards truth and knowledge, the subject actively engages in an avoidance of this very knowledge, so much so that ‘the subject acts not on the basis of what it knows but on the basis of how it desires’ (McGowan, 2013, p. 18). Consequently, for McGowan, what Lacan uncovered was the fact that we are subjects of desire. The world and the objects in it are not simply there to be explored by the subject but are prefigured by the ‘subject’s look’ (2012, p. 41). In so doing, ‘the subject of desire invests itself and thus shapes what it knows; it distorts the apparently external world’ (McGowan, 2012, p. 41).

On this basis, there is no enjoyment in acquiring knowledge – an acquisition that would, fundamentally, extinguish one’s very desire – nor is there the assumption that one can overcome one’s deception in order to reach the truth. Instead, the subject of desire is located in the distortion and deception of the subject itself: what the subject of desire accepts, or acknowledges, is how ‘there is no neutral truth, no truth that doesn’t involve the distortion of subjectivity’ (McGowan, 2012, p. 41). This is not to suggest that we simply shun truth or ignore its importance, but that how ‘one accesses truth [is] through distortion or through deception’ (McGowan, 2012, p. 41), and, perhaps, dissatisfaction.

We can, I believe, witness this distortion and dissatisfaction in our interactions with the AI chatbot (a dissatisfaction that maintains our enjoyment). Although, as a subject of knowledge, we may interact with the chatbot through the satisfaction of acquiring relatively helpful responses to our questions and queries, there is also the underlying sense that the answers provided are never complete. There remains the underlying concern that something may have been missed; that the algorithm may have overlooked a certain possibility; that the answer itself may be wrong and, perhaps, a better question will elicit something else. At no point is truth or knowledge obtained, beyond our own dissatisfaction.

In addition, it is important to remember that Lacan’s (1973/2004) account of desire is one that remains inherently dialectical. Although desire is always the desire of the Other, this does not discount the subject. While desire is not beholden to the subject itself, it is ambiguously located in the subject’s recognition of the Other and their desire. Desire is thus always the desire for what the Other desires. It is in accordance with this account of desire that the ambiguities and complexities in our social relations, both real and online, are produced and maintained.

This is effectively demonstrated in Flisfeder’s (2021) account of the algorithm, where, through our digital relations on various social media platforms, we enact the Other’s desire. Certainly, Flisfeder’s contention should not be perceived as merely describing the desire of the subject for an Other, in this case, social media; nor, should it be conceived as reflecting the desire of the algorithm. Rather, ‘algorithmic desire … is not the desire of the algorithm but is our own, which is also the presumed desire of the big Other’, specifically, the ‘networks of other users’ (Flisfeder, 2022, p. 420). This network is what comprises the online social interactions that underwrite our digital relations and for whom the online user mediates their desire.Footnote 4 While Flisfeder’s argument considers ‘the dynamics between subject-user and the online social network’ (2022, p. 414), importantly, for present purposes, this dynamic can elucidate on the relationship between the subject and AI chatbot, and, more importantly, to an explication on the ethics of desire and drive.

The Ethics of Desire and Drive

For Flisfeder, ‘the ethical subject is’, following Žižek (1991), ‘the subject of drive’, which he summarizes as ‘the subject who has traversed the fantasy and has learned that oneself is the agent responsible for erecting one’s own limits, one’s own barriers to enjoyment’ (Flisfeder, 2022, p. 427). It is for this reason that ‘the ideological subject is … the subject of desire, [the] one who has not yet traversed the fantasy and who still clings to the form of the Symbolic order and the big Other’ (Flisfeder, 2022, p. 427). As evident in the required traversal, the role of fantasy in this account maintains a key importance. Serving as the stage of desire, it is fantasy that ideologically embeds the subject in the pursual of the lost object. Accordingly, ‘in acts of repression or disavowal, for the subject to continue to save itself as a subject of desire – for it to continue following the path of desire – it ultimately displaces the impossibility of the object onto some obstacle that prevents its attainment’ (Flisfeder, 2022, p. 427).

It is possible to compare Flisfeder’s subject of desire with the subject of knowledge: a subject who remains dependent on the fantasy, or the delusion, that one’s knowledge can be obtained in some final truth – a lost object. Whereas the subject of knowledge desires the pleasure and satisfaction in obtaining truth, by linking Flisfeder’s account of desire with the subject of knowledge we can perceive a false desire: one predicated on a dependence on the Other and the possibility of some future access to the unmitigated (lost) object. Further still, in much the same way that the subject erects obstacles and forms of prohibition that sustains its desire, we see a similar form in the acts of disavowal and foreclosure that obscure the gap in knowledge. In either case, what remains important to Flisfeder is that ‘fantasy and desire act as a lure [emphasis added] that nevertheless enables the subject to garner enjoyment in its constant pursuit and failure to attain the lost object’ (2022, p. 427).

While one’s adherence to the fantasy reflects a garnering of enjoyment, in the sense that through pursuing the fantasy of the lost object the subject remains tethered to the symbolic order and the Other, fantasy can also expose that ‘place at which the symbolic order breaks down’, and where the Real is either confronted or experienced (McGowan, 2007, pp. 210–211). Drawing specifically from the films of David Lynch, McGowan elaborates:

The intersection of fantasy and desire is always a point of trauma because it is a point at which signification breaks down. We construct fantasy to cover over a gap in the symbolic structure, a place where there is no signifier. Hence, the hinge that links fantasy to the symbolic structure (i.e. the world of desire) is the real, a traumatic moment that resists all symbolization. (2007, p. 211)

Although Flisfeder (2022) associates the realization that enjoyment resides in fantasy with a move towards drive (itself a traversal of the fantasy), this does not mean that we should so easily avoid the significance of fantasy, as outlined by McGowan (2007), and the importance of desire.Footnote 5 This importance is born out of the (non-)relation that the subject maintains with the Other and its impact upon the subject’s desire. That is:

Although the Other exists as a force binding subjects together, it does not exist as a substantial identity capable of providing a final scale of justice. This nonexistence of the Other leaves subjects on their own when it comes to finding justice and being ethical. Ethics comes down to the subject’s relation to its own desire rather than the achievement of redress or balance in the eyes of a nonexistent Other. (McGowan, 2012, p. 63)

To what extent, therefore, do our relations with the AI chatbot propose or indeed hinder the possibility of an ethics grounded in desire?

For this, we can assert that while Flisfeder (2022) locates our enjoyment in the pursual of desire, enjoyment can also serve as that which defends against our desire (Lacan, 1966/2006, pp. 671–702). This defence proves synonymous of the drive in that ‘drive resolves … [the] endless movement of desire by way of elevating the endless circulation around a lost object into a source of satisfaction’ (Žižek, 2022, p. 235). Here, Flisfeder elaborates on this account of the drive, proposing that ‘what drive enjoys is a constant return to the subject’s foundational act of choosing the self-limit from which its subjectivity was produced in the first place – that is, the choice of a representational signifier’ (2022, p. 426). This choice – this self-limitation – is what can be repressed, disavowed, or foreclosed by the subject (Black, 2023). The (self-)restrictions this creates is transferred to the Other, which remains the prohibiting force of limitation for the subject. In other words, the limit to our enjoyment falls at the behest of the Other’s prohibiting restrictions, rather than the subject’s own choosing. Ultimately, ‘in the act of choosing – that is, of affirming a choice – the subject at the same time negates all the various other possible choices that were previously available to it, and in this way it emerges as a desiring subject’ (Flisfeder, 2022, p. 427). Desire is thus linked to a negation which is deferred to the Other’s apparent restrictions and prohibitions.

It is, however, this desire that remains unconscious to the subject. Here, the negation of the other choices is not consciously affirmed by the subject, but reflective of the desire not to know, so that ‘the existence of the unconscious is the expression of the desire not to know’ (McGowan, 2012, p. 51). In the same way that ‘we can’t be subjects of knowledge because our enjoyment depends on remaining hidden and violating some real or imagined restriction’ (McGowan, 2012, p. 51), equally, it is in the desire for knowledge that such restrictions are fantasized by the subject.

Evidently, while there is a clear distinction between Flisfeder (2022) and McGowan’s (2012) accounts of desire, together, their interpretations can be conceived as a parallax of the ‘gap’ between desire and drive; that is, in each case, ‘desire and drive are the two reactions to this gap: desire externalizes the lack into a cause-object, drive circulates around the object. In desire, the gap appears as lack; in drive, it appears as an excess that derails the circulation of life’ (Žižek, 2022, p. 235). Left in the position of the parallax, all that remains is the revelation of this gap, and it is here that an ethics of desire can be forged – one that is brought to bear in our relations with the AI chatbot.

The ‘Dialectic of Desire’

There is, according to Fink, a move towards a ‘dialectic of desire’ in Lacan that is achieved ‘in the dialectical process of analysis – “dialectical” in the sense that the patient becomes free to say, “well yes, I want that; on second thought, I don’t really; come to think of it, what I really want is …”’ (Fink, 1999, p. 26).Footnote 6 This occurs when ‘the patient no longer feels he or she has to be consistent; he or she can assert a wish during one session, contradict it during the second, reassert it with slight changes during the third and so on’ (Fink, 1999, p. 26). Under such circumstances, the extent to which one’s ‘desire is set in motion’ (Fink, 1999, p. 26, italics removed) is dialectically confronted in the effects of the constitutive lack, in the inconsistency of the signifier, and in the gap inherent to knowledge.

Indeed, the space that this provides is apparent when the confines of desire bear no restriction or prohibition from the Other, but, rather, the Other’s desire, and the ambiguity therein, is fully asserted. Importantly, the dialectic of desire exposes how ‘the unconscious desire (the aim) to achieve a harmonious mesh of object and satisfaction (the goal) only reveals [emphasis added] a gap between the goal and the object’ (Ragland, 1995, p. 197). In fact, whereas ‘desire is grounded in its constitutive lack, [and] while drive circulates around a hole, a gap in the order of being’ (Žižek, 2006a, p. 61), it is in the case of desire that lack – the gap – is revealed: a revelation that goes beyond drive’s endless circulation.Footnote 7 Such an emergence or revelation of desire remains integral to the analytical process, and to the revelation that our chatbot relations prefigure.

Faced with the AI chatbot, what remains apparent is the way in which our various interactions, and perhaps academic interventions, disclose a desire not to know. Importantly, what the desire not to know reveals is a certain space, a point at which our desire is not predicated on our desire not to know our own desire, but, rather, in the revelation that it is the desire not to know the AI’s desire (the Other’s desire) that is exposed. In this regard, the ‘stupidity’ of the AI chatbot – including its various mistakes, inaccuracies and failures – does not inhibit our interactions with the AI (if anything, such mistakes can encourage our further interaction). Instead, such inconsistency in the AI Other and its advertence of lack, function as a defence measure that inherently performs the desire not to know. Through enacting such a defence, the desire not to know is constituted in the denial of truth and the distortion of the subject – a distortion that is brought to bear in the chatbot’s various mistakes. Although our desire not to know remains unconsciously invested in our turn to the AI as a source of knowledge, it is in the intersection of lack between the subject and the AI Other that the centrality of loss is affirmed, and the distortions of the subject revealed.

This sheds new light on the well-trodden assertion that to ‘hav[e] given ground relative to one’s desire’ (Lacan, 1986/1997, p. 319), is, in part, to remain dependent on the knowledge that is assumed in the Other’s desire (that is, our desire remains tied to the Other’s desire). Rather, in accordance with the AI chatbot, what the desire not to know prescribes is the opportunity to reveal the singularity of our desire. As Ruti explains, ‘our responsibility, in a way, is to singularize our desire – to stop paying attention to what the Other wants and to focus instead on what we ourselves want – so as to avoid being completely subsumed by the desire of the Other’ (2012, p. 50).

Such assertions allow us to better approach the AI and, in particular, the uses to which it is put. Indeed, whether used to help answer a particular inquiry or to help draft a required text, our turn to AI reveals the potential to embrace the very loss it avers. If our use of AI goes no further than a simple characterization, which views our playful excursions with AI as nothing more than a pointless endeavour to create more knowledge, true or false (Dean, 2009), the fact that our desire can never be satisfied lays open a path that acknowledges the centrality of loss for the subject, and, more importantly, to the role this loss plays in our turn to AI and the (dis)satisfaction this procures. A space, perhaps, where our online satisfactions, however banal (Dean, 2009), are not sustained by a knowledge that seeks to disavow loss, but which instead better recognizes and appreciates the potential in loss itself. This, I believe, will remain an integral component of our future relations with AI. In light of its mistakes, AI technology will undoubtedly improve, yet these mistakes and failures will not disappear. Indeed, to assume otherwise would be to succumb to a paranoid conviction that the AI will achieve or can achieve a ‘total knowledge’, thus becoming an entity devoid of lack. More to the point, such fantasies do not necessarily ignore the importance of desire. Outside of the very fears and anxieties that underscore our adoption of the AI chatbot is the opportunity to render a transformation in our digital lives. In this sense, the desire not to know reveals the opportunity to assert and define the gap inherent to both the subject and the AI we create.