AI, Automatons, and Modern Insanity
E. T. A. Hoffmann was a man of his age. A lawyer by day and a writer by night, Hoffmann wrote stories that are typical of German Romanticism: uncanny and unsettling; they explore insanity, the supernatural, and man’s struggle against forces more powerful than himself. His life, too, fits the stereotype of the tortured Romantic poet. Like one of the strong-willed characters in his stories, Hoffmann was unable to control his impulses, and his fiction sometimes damaged his prospects in the real world. Alcoholism and syphilis led him to his early death at the age of 46. It may seem unlikely that a man so of his era would have much to say to our own time. Yet, his fiction is strangely prescient. In particular, his story “The Sandman” anticipates the development of artificial intelligence in the modern world and offers us useful ways of considering the potential consequences of AI today as well as our entanglement with it.
“The Sandman” tells the story of a young college student named Nathaniel. It begins with a letter to Lothario, his fiancée Clara’s brother, in which he confides a traumatic childhood experience of being terrorized by one of his father’s associates, a man named Coppelius. As a boy, Nathaniel believed Coppelius was the legendary “sandman” who steals the eyes of children. Now, years later and miles away, Nathaniel is certain he has seen Coppelius again, and he becomes so obsessed with the idea that he has to leave school. Clara and Lothario reassure him that his fears are all in his head, the wild exaggerations of a child’s imagination.
But just as Nathaniel seems to regain his mental equanimity, a new disturbance overtakes him. He falls in love with a professor’s daughter, Olympia. Initially, Nathaniel only catches glimpses of her from afar, but he finally has the chance to meet the shy, nearly mute girl at her social debut. Olympia is Nathaniel’s ideal. She sings, dances, and plays the piano. Whenever they are together, she listens to him for hours on end without interrupting. Whereas the practical Clara would check his flights of fancy, Olympia always indulges and affirms him with a simple, “Ah, ah.” At first, Olympia’s eyes seem “strangely fixed and dead,” but they quickly grow “ever warmer and more lively.” She has little to say, but he will settle for her adoring gaze and always available hand, waiting to be warmed by his.
His fellow students regard the new love with some skepticism. One asks how “a clever chap like you could possibly have been smitten by that wax-faced wooden doll over there?” He continues, “we have come to find this Olympia quite uncanny; we would like to have nothing to do with her; it seems to us that she is only acting like a living creature, and yet there is some reason for that which we cannot fathom.” Nathaniel responds that others simply cannot comprehend the depth behind her few words, which “appear as genuine hieroglyphics of an inner world full of love and a higher knowledge of the spiritual life in contemplation of the eternal Beyond.”
Before long, Nathaniel isolates himself from all others and resolves to marry Olympia. But his plans are interrupted when he discovers a terrible truth: Olympia is not a person at all—she is an automaton. Nathaniel discovers the truth in the worst possible way when he sees Olympia being carried off without her eyes. The professor cries out that Coppelius has robbed him of his “finest automaton.” The revelation shatters Nathaniel, who begins yelling out broken fragments from a childhood song—“Spin, puppet, spin!”—and nearly kills the professor in a fit of madness.
Speculative cross-section of Jacques de Vaucanson’s 1974 Canard Digérateur automaton
Nathaniel returns home to his family and fiancée, who nurse him back to health. One day, he and Clara are enjoying a scenic view from a tower, and Nathaniel believes, once again, that he has spotted Coppelius. Nathaniel begins screaming, “Spin, spin, circle of fire! Spin, spin, circle of fire!” and nearly kills Clara before throwing himself off the tower. Below the tower, Coppelius laughs at Nathaniel’s plight before slipping away into the crowd.
Like today’s AI, the famous automatons of early modern Europe seemed to do the impossible, and they foreshadowed a future in which humans and technology could become indistinguishable. The sixteenth-century mechanical monk of King Philip II nodded and kissed the rosary and appeared to pray. It was considered a “miracle.” The famous eighteenth-century Canard Digérateur, or “Digesting Duck,” was called the “glory of France.” Other automatons could play music, like the Marie Antoinette automaton that plays the dulcimer. These early automatons were marvels and took beholders to the uncanny valley, much like technology and artificial imagery do today.
Hoffmann’s story “The Sandman” reflects a deep concern with automatons, and its depiction of the life-like Olympia is prescient about AI. Some of Nathaniel’s friends found her “uncanny,” but only the “extremely astute students.” Others marveled at her abilities, much like many people marvel at today’s large language models (LLMs) and are unable to discern their work from the results of human efforts. Both AI and Olympia were designed to please. Olympia performed, listened, and affirmed. Many men envied Nathaniel the perfectly attentive, beautiful woman he had secured. LLMs are so eager to please that they will sometimes invent the sources you need to make effective arguments in your research papers or legal briefs. ChatGPT is notorious for generating fake sources. And while LLMs will sometimes acknowledge limitations, other times they will “hallucinate” outputs that do not make sense. As Olympia says, “Ah, ah.”
E. T. A. Hoffmann found the idea of falling in love with an automaton nightmarish, but it could soon be our new normal. Today, it is becoming more common for people to encourage emotional and romantic connections between humans and AI. We are already living in the world shown in the 2013 Spike Jonze film Her. All kinds of AI companions are now available—girlfriends, therapists, trainers, mentors, nutritionists—sometimes for free. A New York Times columnist who test drove 18 AI companions for work confessed, “I’ll be honest: I still vastly prefer my human friends to my A.I. ones…But on balance, they’ve been a positive addition to my life, and I’ll be a little sad to delete them when this experiment is over.”
It is likely that the trend will only increase in the next few years. That same column continues:
I expected to come away believing that A.I. friendship is fundamentally hollow. These A.I. systems, after all, don’t have thoughts, emotions or desires. They are neural networks trained to predict the next words in a sequence, not sentient beings capable of love.
All of that is true. But I’m now convinced that it’s not going to matter much.
The technology needed for realistic A.I. companionship is already here, and I believe that over the next few years, millions of people are going to form intimate relationships with A.I. chatbots. They’ll meet them on apps like the ones I tested, and on social media platforms like Facebook, Instagram and Snapchat, which have already started adding A.I. characters to their apps.
“The Sandman” suggests how quickly we might be able to adjust to the reality of “hollow” AI companionship. Even Nathaniel initially notices Olympia’s empty eyes. The eyes are “the window to the soul,” and hers are entirely impassive. Yet, Nathaniel soon learns to overlook those “lifeless” eyes. Indeed, the story seems to confirm that although AI only simulates human thought and emotion, it might not “matter much.”
The dramatic ending of “The Sandman” challenges the optimism of AI enthusiasts. Neither Olympia the automaton nor AI are neutral or objective creations. Olympia was designed by someone, and she does not think or operate on her own. Even though the LLMs we have today may appear to learn on their own, they do not have their own consciousness, and they bear the fingerprints of their makers in their programming. We might think of AI as a tool or Olympia as a puppet, but the raving Nathaniel shows another possible outcome: Olympia has made a puppet of Nathaniel. He is not in control. He is being controlled by Coppelius. The risk we run with AI is that as we become more dependent on it, we will become more subject to it and its creators.
We can already see some ways in which AI is capable of reducing us to puppets while it pulls the strings we have given it. AI chatbots have led to suicide in the real world. Psychologists are warning about the dangers of falling for an AI girlfriend. According to Wired, “AI girlfriends are a privacy nightmare.” The same is true of AI therapists. AI companions can serve us up to Big Data and sell us to advertisers. As Nathaniel cries out, “Spin, puppet, spin!” Rather than making us more comfortable and at home in the world, AI companions can pull us deeper into a digital reality where we and our data are the currency, and we are more distanced than ever from the actual people and world around us.
Karl Gottlieb von Windisch, Schachtürke (1784)
When Clara and Nathaniel’s friends try to pull him back from the edge of insanity, they encourage him to take control of himself and the situation. Early in the story, Clara is certain that the Coppelius from Nathaniel’s childhood was not really a monster or capable of harming him, but rather the product of Nathaniel’s imagination. Clara’s brother warns Nathaniel “that this dark psychic power, once we have surrendered to it, often assumes other forms which the outer world throws across our path and draws them into us, so that the spirit which seems to animate those forms has in fact been enkindled by us ourselves.” While Nathaniel doubts his own free will, others claim that the only real threats to Nathaniel come from within and if he has proper mastery of himself, he’ll be fine. Clara maintained that “Coppelius is an evil, inimical force, he can do terrible things…but only so long as you fail to banish him from your mind…his power is only your belief in him.”
Like Clara, proponents of AI argue that as long as we are masters of ourselves, we needn’t worry that AI will master us. But as all the writers of the Romantic era knew, men are helpless when in the thrall of powers greater than themselves. AI may not be as sublime as a Romantic seascape, but it is at least as vast. Even if it were easy to resist AI, right now it is almost impossible to avoid it. AI (or something called “AI”) has been inserted into your web browser, your email, your word processing software, your medical diagnoses and care plan, and your insurance coverage. Slipping into dependence on AI is easy and will become easier if it lives up to its earning potential. As the end of “The Sandman” suggests, self-mastery may not be enough. It will take more than confidence to prevent ourselves from being turned into puppets and from losing the power of our own eyes to tell the difference between the real and the artificial, between life and what is lifeless.
Elizabeth Stice is an associate professor of history at Palm Beach Atlantic University. She has a book about World War I, Empire Between the Lines: Imperial Culture in British and French Trench Newspapers of the Great War, and she has written for various publications, including Front Porch Republic, Comment, and Inside Higher Ed.