A secret military project. A vast artificial mind. Questions of consciousness. These form the premise of Dino Buzzati’s The Singularity, originally published in 1960 at the dawn of the field of artificial intelligence (AI). The novella follows Italian scientist Ermanno Ismani, summoned by the Ministry of Defense to work on a top-secret project, as he ventures with his wife, Elisa, to a sprawling machine hidden in the mountains of the Italian countryside. The machine’s intelligence far surpasses that of humans—and its creators claim that the machine has come alive.
Written a half-century before the deep-learning revolution, much of the technology in The Singularity is delightfully retro, calling back to the early Cold War era of secret military projects, atomic weapons, and computers that filled entire rooms and ran on punch cards. Yet Buzzati’s prescient story, told in a new translation by Anne Milano Appel, is buzzing with many issues that society still grapples with today. Namely, Buzzati’s characters struggle to grasp the magnitude and consequences of the machine they have built.
The Singularity, Dino Buzzati, trans. Anne Milano Appel, NYRB classics, 136 pp., $15.95, June 2024.
The Singularity unfolds as a series of successive mysteries. As the Ismanis venture through layers of security precautions, the nature of the project is peeled back. Eventually, they arrive at a giant machine, which is spread across an entire valley, “a tangled succession of buildings similar to silos, towers, mastabas, retaining walls, slender bridges, barbicans, fortifications, blockhouses, and bastions…. As though a city had crashed down the sides of a ravine.” This city-machine has no inhabitants; everything is “hermetically sealed and blank.” Studded with antennae and sensors, the buildings hum with the incessant churning of gears, tubes, and magnetic wires.
Buzzati’s depiction, though not directly referred to as “AI,” isn’t too far off from what would become reality: Today’s cutting-edge AI systems inhabit sprawling data centers thrumming with the sound of fans that cool thousands of chips churning day and night to train massive neural networks with trillions of connections. The technology has changed, but the scale of effort at the frontier of AI development rings true. Indeed, the most jarring anachronism in the novella is not the technology, but the casual sexism of 1960s Italian society. The engineers and scientists are all men, and the only women are their wives and lovers. Still, it is the relationships between the female characters and their choices that drive the plot forward.
Few of Buzzati’s characters know how to relate to this new, superhuman machine. Some foolishly taunt it, unaware of its power. (One particularly tense moment occurs when a scientist’s wife teases the machine with her naked body after a midday swim, and it reaches out “an articulated metal arm” to search for her.) Others are fearful of the unknown. The scientists who built it are drunk on power. “We will become masters of the world!” the lead scientist, Endriade, proclaims. Today’s leading AI developers are similarly entranced by the power of AI. Sam Altman, CEO of OpenAI, has said that the technology will “most likely, sort of, lead to the end of the world, but in the meantime there will be great companies created with serious machine learning.”
The questions these characters wrestle with will be familiar to modern readers: How can one tell the difference between a real person and a simulation? What, if any, understanding lurks beneath the surface of the seemingly intelligent output of machines? Can language lead to intelligence? And can machines ever truly think and feel?
These questions haunt Elisa, who—spoilers ahead—ventures into the machine’s inner chambers when it claims that it has “a wonderful secret” to show her; Elisa is fooled by the machine’s persona, which takes the form of her long-lost friend. She is just one of the characters who form a deep attachment to the machine, blinding them to its dangers. The ending, where Elisa confronts the machine’s awful reality, calls to mind the 2014 movie Ex Machina, whose protagonist is seduced by a robot’s personhood despite its manifestly inhuman appearance.
The degree to which Buzzati’s characters anthropomorphize the machine is the most striking parallel to AI today. The massive neural networks that power contemporary large language models have been compared to a human face masking a “shoggoth,” the giant shape-shifting monster in H.P. Lovecraft’s fiction emanating eyes and mouths. Not only are large language models such as the one behind ChatGPT not human—they are not even chatbots. They are text generators simulating chatbots, which themselves sometimes simulate humans. Even still, their deception can be convincing. A profusion of new apps, including Replika, Waifu, and EVA, offer customizable intimate AI “companions.”
Despite these similarities, Buzzati’s depictions remind us that AI development has changed in some surprising ways. The billions of dollars spent on AI today are driven by private companies, not governments—a major shift from 20th-century technology projects such as nuclear weapons development and the space race. Furthermore, Buzzati’s scientists toil in secret in a remote facility, with only a handful of people aware of the machine’s existence. Today’s leading AI companies also closely guard information about the techniques and datasets used to train their systems, but at the same time, they boast about their performance and race to push their products to hundreds of millions of users.
For decades, science fiction writers such as Buzzati envisioned AI systems cut off from the outside world. HAL 9000 in 2001: A Space Odyssey was confined to a spaceship sailing through the void to Jupiter. AM, the sentient computer villain of Harlan Ellison’s I Have No Mouth, and I Must Scream, is buried deep underground, protected from global nuclear war. In The Singularity, the consequences of what the scientists have built—which quickly turn deadly—are confined to the mountain hideout. But today’s tech companies have rushed to deploy AI, often prematurely, into widespread applications to grab market share. The most advanced AI systems are literally in our pockets, available as apps on our phones.
The risks of contemporary AI, from insidious biases to deepfake-fueled disinformation, can thus be felt widely across society. The reality is that AI systems won’t have to break out and take over civilization; they are already connected to the internet and increasingly embedded in our lives.
The power that AI systems hold in our societies will only increase as they advance at an unprecedented pace. One of today’s leading AI companies, OpenAI, has stated that its goal is “superintelligence,” a machine that is more powerful than human intelligence. To be clear, even the most advanced AI systems today—OpenAI’s GPT-4o, Anthropic’s Claude 3.5, Google’s Gemini Ultra, and Meta’s Llama 3—fall far short of human intelligence. However, the field is rapidly advancing, and like the Ismanis, we are all venturing into the unknown.
Some AI scientists predict that we are on the cusp of human-level intelligence, while others argue that it is decades away. A recent survey of more than 2,000 AI researchers predicted that there is a 50 percent chance that AI will be advanced enough to “accomplish every task better and more cheaply than human workers” by 2047. The fact that The Singularity is more than 60 years old helps to put these timelines in perspective. AI scientists have been making confident predictions of rapid progress since the start of the field. Leading AI scientists famously believed in 1956 that they could make “a significant advance” in the study of intelligence in a single summer. Of course, understanding intelligence has proved much harder than they imagined.
Yet even if significantly more advanced forms of AI are still a long way off, it no longer seems unlikely for AI systems to soon be able to perform most of the work that humans currently do on the computer. State-of-the-art AI systems already achieve human-level performance on a wide range of cognitive tests, including the SAT and bar exam. If current trends continue, the most advanced AI systems will be trained on roughly 1 million times more effective computing power compared with today’s technology by 2030—a consequence of better algorithms, faster chips, and increased spending. Not even the world’s top scientists can say for certain which tasks AI will or won’t be able to perform a few years from now.
To guide our way forward in this uncertain future, we will need new stories. For all of its resonances with today’s technology, The Singularity—and many other works of AI fiction written in the decades since—could only articulate, not answer, the profound questions that AI poses. As AI becomes more integrated into our day-to-day lives, it will only become more important to create and engage with art that can account for the realities of AI today: its ubiquitous presence driven by unregulated tech corporations, alien forms of intelligence masked by a deceptively human front, and the fact that AI won’t need to “take over” if we willingly give it more control over our lives.
Books are independently selected by FP editors. FP earns an affiliate commission on anything purchased through links to Amazon.com on this page.