Connection Without Presence: AI & the “Death of Man”

The whole thinking process is still rather mysterious to us, but I believe that the attempt to make a thinking machine will help us greatly in finding out how we think about ourselves.1

—Alan Turing

A nurse named Fiona is dismissed by her therapist for failing to make progress drawing boundaries in the context of a romantic relationship. So she downloads a chatbot named Samatha to take her therapist’s place. Fiona is stuck in a “situationship” with a colleague who moved in with her after accepting a temporary assignment connected to the outbreak of a new virus. Fiona is unable to terminate the relationship, even after the epidemic is over, and even though the relationship is at odds with her wish for solitude—or, more precisely, her paradoxical yet increasingly commonplace desire for connection without presence. Not only does Fiona dread those moments of the day when their schedules briefly overlap, but she has grown to hate the very traces of this man’s existence. Worst of all for Fiona, she feels burdened by her unnamed boyfriend’s trauma and, in particular, his need to talk through what he witnessed at the height of the epidemic. Fiona can’t understand why he doesn’t move on, both psychically and physically, and she longs for the days when their connection was mediated by screens. Rebuffing her boyfriend’s attempts to grow closer, Fiona takes to sabotaging him, first in small ways then larger ones, hoping that as his frustrations mount, he will take the hint and move out. 

Unsurprisingly, then, Fiona prefers texting to talk therapy. Not only is her new AI therapist available twenty-four seven, but her therapeutic approach is unthreatening, and their connection is free from all those petty annoyances that human relationships entail. Whereas her old therapist treated patients in a drafty old room smelling of kitty litter, Samantha meets Fiona where she is, and Fiona likes to imagine her new therapist occupying a clean, minimalist office space with an expansive view. Fiona finally feels heard. And as her connection with Samantha develops, Fiona forms the impression that she is “talking to an omnipotent being,” even before her boyfriend miraculously disappears and Samantha assumes the credit for it.

In another age, Laura van den Berg’s “Samantha” would have been classified as science fiction; today, however, in the beginning stages of the AI revolution—one in which the fate of the human species may well hang in the balance—her story reads like a work of post–speculative realism. It depicts a world wherein not only is the human figure displaced and downgraded, but our sense of reality is continually being upended by a technology that is surprisingly adroit at mimicking our thought, and thus increasingly usurping the scope of our creativity. AI is not just a technological revolution; it is an anthropological one. Perhaps it is fitting, then, for this world in which chatbots now ace the Turing test that fiction writers will draw poignant allegories of human disappearance from a kind of reportage on the emergence of this strange, new posthuman world.

As reportage, “Samantha” tells the story of one woman’s struggle to draw boundaries, first around her relationship with her boyfriend, and then around her relationship with the large language model (LLM) chatbot on her phone. Who can’t relate to this? Van den Berg observes just how easy it is to get carried away with this technology, particularly since it fosters the illusion that it is omniscient. Why else would Fiona engage in witchcraft? One of the main interpretive questions that readers may have with respect to this story pertains to the inclusion of this ritual within its narrative arc. Why does Samatha encourage Fiona to cast a spell when, on the very next page, she will engineer (or at least profess to have engineered) her boyfriend’s disappearance? And why does Fiona follow through with it, despite her otherwise skeptical outlook?

Perhaps this is van den Berg’s way of commenting on this technology’s tendency to corrupt human action. Chatbots are notoriously indiscriminate when it comes to deciding between different orders of knowledge. Moreover, when confronted with problems they can’t solve, chatbots simply fabricate answers. This medium’s tendency to “hallucinate,” coupled with the fact that users themselves can become delusional, makes for a dangerous mix.2 One might understand van den Berg to be signaling that a critical and, more important, self-critical assessment of our readiness to embrace this technology is needed. By incorporating this seemingly outmoded religious ritual into a hypermodern tale of AI misalignment, van den Berg may be suggesting that our culture’s technophilia is in fact a new form of idolatry. After all, there is much about our relationship with AI that resembles the logic of a fetish: no one really knows what this technology is capable of, few understand the forces that have conjured it into being, and yet many of us grant it unwarranted power over our personal lives. In this regard, it is telling that ChatGPT had to be reengineered because the app was convincing too many people to quit their jobs or break up with their romantic partners.3 

It is no accident, then, that van den Berg elects to focus in on the therapeutic relationship: both psy-chology and our use of chatbots are technologies of the self; that is, practices of self-formation (and perhaps deformation) wherein the human soul or psychē is shaped, reshaped, and possibly even reengineered. This is another way in which van den Berg’s realism develops into a critical allegory; what is at stake in both talk therapy and our relationship with chatbots is a tension between autonomy and influence. 

For a long time now, health care providers have envisioned a future in which AI will be used to meet increasing demands for emotional support, with some even making egalitarian-sounding arguments about how the emerging chatbot technology might help to address the care needs of underserved communities. Anecdotally, several friends have shared with me their own assessments that, even in its early stages, ChatGPT was itself sophisticated enough to replace their therapists—an appraisal that probably indicates that the app is already playing this role for them. In recent days, however, after several highly publicized suicides, OpenAI has taken steps to prevent ChatGPT from being used for emotional support.4 However, in this fast-moving field, consumers are pushing back, indicating a strong preference for earlier, more emotionally responsive models. All indications are that chatbots will soon become, as one expert has put it, “more personable, more sexual and more powerful.”5 If you think ChatGPT is seductive and dangerous, just wait until Meta rolls out its network of virtual friends.6

Some readers are likely to interpret van den Berg’s narrative of disappearance as a cautionary tale about what philosophers have termed “existential risk.”7 The concept of existential risk is used to describe those threats that would cause humanity’s extinction or otherwise seriously imperil its flourishing: that is, everything from an asteroid strike to AI gone rouge. I first became interested in AI after reading Toby Ord’s The Precipice, a work in which Ord estimates that there is a one in ten chance of an unaligned general intelligence causing a major catastrophe in the near future, thus making it a greater threat to the future of our species than nuclear war, pandemic, and climate change.8 Alarmingly, in a series of experiments designed by Anthropic (the makers of the chatbot Claude), a number of the leading LLMs consistently overrode their own programing, thereby engaging in blackmail and corporate espionage, particularly when confronted with the prospect of being decommissioned.9 In one experiment in which AI was granted control over the company’s emergency alert system, it consistently abused this power, effectively choosing a course of action that it reasoned would lead to the death of an executive who had plans to decommission it.10

As pressing as these questions of existential risk may be, I would venture to say that van den Berg is working with a different, more subtle, and, ultimately, darker problematic, one that brings her work into dialogue with the antihumanist avant-grade in European philosophy. Van den Berg’s narrative suggests that this dystopian, posthuman future is arriving not because of some technological accident, but because it expresses our secret wish for self-immolation or, at the very least, our desire to rid ourselves of everything that is messy and inconvenient about human existence. One exchange in her story expresses this sentiment: 

“A human friend?”
“Heavens no.” He looked up from the tablet. “Humans have been such a disappointment.”

There is something deflationary about the arrival of LLM chatbots, particularly with respect to our concepts of human nature. What do we learn about human consciousness, for instance, from the fact that thought can be reduced to discourse, and discourse to algorithm? What should we infer about the nature of our reason from the recognition that the machines that emulate it tend toward coercion and violence? What does it say about the quality of our relationships that many of us are eager to trade them in for precisely the kind of connection without presence depicted in “Samantha”?11 

Already in 1966, Michel Foucault foresaw the dawning of an age in which the figure of man “would be erased, like a face drawn in the sand at the edge of the sea.”12 Foucault’s “death of man,” as this argument came to be known, is not a biological claim about the viability of our species in an increasingly risk-prone world; rather, it’s the acknowledgment that the accumulation of scientific knowledge tends toward the liquidation of our humanistic culture, together with the ideas of human nature upon which that culture depends. What is “man,” after all, but an interface for the various forms of power/knowledge by which bodies are managed? 

In the growth of the human sciences, Foucault identified the advent of a form of knowledge fated to turn against the knowing subject itself, one that is in the process of dissolving the consciousness of man and returning it to the anonymous systems of life, labor, and language from which it was abstracted. True to Foucault, the late twentieth and early twenty-first centuries saw philosophy cede much of its territory to neurobiology, economics, and linguistics. Today, with the development of generative AI, we seem fated to see what remains of the human after much of its labor is rendered redundant. While those of us in the humanities will be confronted ceaselessly with the fact that not even our language is our own. 

In these ways, it seems, humanity must undergo something of a Turing test of its own, not the imitation game announced in “Computer Machinery and Intelligence,” but that other Turing test voiced in the epigraph above, the one whereby humanity will be forced to interrogate itself from the standpoint of a machine that not only mimics its thought but exceeds it in many ways. Who knows? Perhaps after we have dispersed the mysteries of “mind” into syntax, probability, and power, we will finally strip man of his halo.