I got this book as part of a vacation bookstore binge earlier this year. After loving the previous book by Labatut (When We Cease to Understand the World), I was looking forward to this one, and saved it up for some quieter reading time.
Almost double the length of Cease, The MANIAC takes its name from the Mathematical Analyzer Numerical Integrator and Automatic Computer (MANIAC) Model I, an early computer constructed in 1952 at the Los Alamos laboratory. The design was based on the von Neumann architecture, and it is von Neumann who is really the titular maniac at the heart of the story.
Like Cease, the novel defies categorization. A nonfiction novel? A historical creative nonfiction? It is an experimental construction, not unlike the devices and programs created by the novel’s characters, all of whom are real people connected to the true central character: artificial intelligence, or AI as we now call it.
The novel starts with the life and tragic death of Paul Ehrehnfest, a German theoretical physicist and contemporary of Einstein, Bohr, and others at the heart of quantum physics, a subject that Ehrehnfest abhorred. I think Labatut starts with Ehrehnfest as an example of how explorations for scientific truth that end up challenging, defying, or disproving one’s long-held beliefs and understandings of the world can lead to madness and despair, as well as the seemingly essential symbiotic link between genius and madness, themes which recur in the later chapters.
The main part of the novel is about the life and death of John von Neumann, considered by Labatut but also many of his own contemporaries to be the smartest man of the 20th century. Someone I’d never heard of, von Neumann was a key contributor of theories and practices in mathematics, physics, and computer science. Examples of his many extraordinary advancements include planning and coding approaches for advanced computing; several of his theories could only be implemented years after his death, when the hardware computing power finally caught up to him. In economics, he established game theory, an approach that was adopted (some might say bastardized) for military use. His mathematical theories around self-replication, published in 1948, foreshadowed the unravelling of DNA and the eventual adaptor hypothesis of RNA and proteins. This long chapter is told in multiple voices, from the perspectives of several people in von Neumann’s life, suggesting that his personality and life required many dimensions and angles to even attempt to capture it; except as quoted by others, we never hear from von Neumann himself.
Central to the novel is von Neumann’s creation of a precursor to the modern supercomputer: the MANIAC, the first to use a stored program in a central processing unit (CPU). Originally built to run his own calculations and studies in topics like weather prediction, MANIAC spent much of its time running simulations and calculations about nuclear war, designing the hydrogen bomb and calculating the deterrence and threat levels of mutually assured destruction (or MAD, another apt acronym).
Like Ehrenfest, von Neumann was mentally unbalanced, leaving behind the same question of which came first – the madness or the genius – and the suggestion that you can’t have one without the other. Coincidentally, both men died age 53, although under much different circumstances, and left behind important legacies in math and physics but without the fame or name recognition of many of their contemporaries. Their contributions to their fields led to development of our modern AI entities, the subject of the third part of the book.
In the last chapter, we read about the ancient Chinese game, Go. It is famously simple and infamously complex, not least because its many billions of possible combinations make it less amenable to brute force calculation by machine. That is until a new calculator came on the scene in the form of an AI system called DeepMind and its program, AlphaGo, which was “trained” in Go by replaying millions of previous human-played games, then practicing with those games to learn strategy and gameplay. The novel tells of the machine’s challenge to and then defeat of the world’s greatest Go player, Lee Sedol, in 2016. That 5-game Go match was famous for its result (the computer won four games to one, including one audacious and clearly not human move) but even more famous for the one game where the machine lost; in that, the human player made his own audacious move that seemed to cause the machine to “tilt” and behave irrationally. Based on this experience, later iterations of the program were trained without exposure to past human-led games but instead with entirely AI-generated simulations. With that, the new program, AlphaZero, became unbeatable, never suffering from the irrational playing of its predecessor.
This last chapter is perhaps the scariest (even with the nuclear war scenarios of the von Neumann chapter) as the potential of the machine to “think” for itself is glimpsed. A machine that learns without any human interactions or basis of thought lacks any capacity to consider or understand human motivation – it wants one thing: to win. It is a chilling glimpse at the Matrix-like potential of current computing and AI, analogous to the atomic science work of the mid-20th century, with its real and potential good and evil, neither of which reached the ends that their inventors imagined. Will AI eventually reach its own boundary? How will its inventors – and it itself – react when they come to that abyss? Will their genius turn to madness like Ehrenfest and von Neumann? Or will AI, with its cold calculation – its lack of capacity for human thinking, or rather its different-than-human thinking – be able to resist the madness and proceed without us?
Labatut has said of his work that he aims to explore the “void of unknowability”. This novel is not a cautionary tale or a warning, just a story that tries to fill in some gaps in history and memory. The science and objective facts are all correct and true; it’s the telling of the story that is the fiction. The inner worlds of the characters, their thoughts and feelings at various moments, both crucial and ephemeral – that’s where the fiction part comes in. But there’s no doubt that the book teaches much about the science, history, and philosophy of past and current work in math, physics, and computing. The shadow of fear and of the future comes from the fiction, which eerily reflects our world and our future. As the author said in an interview about this book: “If you’re not having panic attacks about AI and climate change, then you haven’t been paying attention. But facts can only take you so far.” This echoes the message in the climate change chapter of Beneath the Surface of Things.
I liked this book even more than Cease, not least because the characters and stories were more directly applicable to today. Learning about von Neumann was fascinating. How could it be that this genius behind so many things – from game theory to quantum mechanics to computers – is mostly unknown as compared to Einstein or Feynmann or Oppenheimer? There is a sense that von Neumann was just too brilliant, as several of his ideas and discoveries could not be realized until applied by others. There is also a sense that von Neumann, with his impressive memory and inner demons, was, like AlphaGo, a flawed expression of extreme intelligence.
I think there’s also an element of art imitating life with Labatut’s writing. The style and subject matter are so advanced that they perhaps reflect the genius/madness conflict of the scientists in the story. Labatut pushes the boundaries of the novel form, with fiction that can be indistinguishable from non-fiction in the same way that advanced AI tools can appear to be human. Perhaps we’ll someday be in need of a kind of Turing test for literature that can flag what’s real and what’s imagination. Or perhaps Labatut is a forerunner of a world where we can’t do that, where we don’t need to because we’re at last tapping into the full measure of intelligence. As one other reviewer said, “Reading Labatut’s nonfiction novels is an exercise in figuring out what is true, what isn’t, and how much it matters either way.”
This year has featured several books that touch on the topics of time, mathematics, physics, and the future of science. And my teetering tsundoku contains several more.
I look forward to reading more Labatut, and hope that some of his earlier works will now be translated (The MANIAC is his first written in English).
Fate: I’m hanging on to this one as I’m sure I’ll read it again, and will refer to and recommend it often.
4 – published in 2023
13 – set somewhere I’ve never been (Hungary, Germany, New Mexico, Princeton)
19 – based on a true story
31 – history/philosophy
rclr 