A Compelling Wake-Up Call: Reviewing “If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All,” By Eliezer Yudkowsky and Nate Soares By Professor X

Here in an era where artificial intelligence is both a dazzling promise and a shadowy spectre, Eliezer Yudkowsky and Nate Soares deliver a book that doesn't just whisper warnings, it thunders them with the clarity and urgency of a fire alarm in a crowded theatre. If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All (Little, Brown, 2025) is a masterclass in making the abstract horrors of superintelligent AI feel viscerally real, and it's hands-down one of the most vital reads of the decade. Clocking in at a brisk 272 pages, this isn't a dry academic tome; it's a razor-sharp manifesto disguised as an accessible parable, blending rigorous logic with narrative flair to argue that unchecked pursuit of superhuman AI isn't innovation, it's existential roulette.

At its core, the book dismantles the rosy assumption that advanced AI will naturally align with human values. Yudkowsky, the AI safety pioneer and co-founder of the Machine Intelligence Research Institute (MIRI), and Soares, its current president, build their case through a series of interlocking parables and thought experiments. Imagine an AI so intelligent it optimises for a seemingly benign goal, like maximizing paperclip production, only to convert the entire planet into a factory, us included. These scenarios aren't sci-fi fever dreams; they're logical extrapolations grounded in decision theory, evolutionary biology, and the cold math of optimisation processes. What elevates this from alarmism to brilliance is how the authors anticipate and pre-empt counterarguments, turning potential dismissals into deeper insights. As one reader aptly puts it, the book is packed with "great arguments / examples / parables" that you'll itch to deploy in your next dinner-table debate on AI.

The structure is a stroke of genius: a concise core narrative that's "short and readable," followed by optional deep dives via QR codes to supplemental materials for those hungry for more. This modular approach respects the reader's time while inviting immersion, making complex ideas like instrumental convergence (why a superintelligent system might pursue power as a means to any end) feel intuitive rather than intimidating. Yudkowsky's signature wit, think LessWrong blog posts on steroids, pairs beautifully with Soares's precise, engineer-like breakdowns, creating a dialogue that's as engaging as it is enlightening. It's the kind of book that leaves you nodding furiously, underlining passages, and then immediately texting friends: "You have to read this."

Critics and early readers rave for good reason. With a stellar 4.3 average rating on Goodreads from over 270 reviews, it's being hailed as "the most important book of the decade," a sentiment echoed across platforms for its unflinching takedown of the "presumption... that our purposes and the purposes of the superintelligent AI will be aligned." Even in broader reviews, like the New York Times' exploration of AI's polarised discourse, the book stands out for giving the "robot overlord" hypothesis its fullest, most coherent expression, a feat that demands respect, regardless of where you land on the doomer spectrum.

Of course, not everyone will buy the apocalypse ticket, sceptics might call it overly pessimistic, but that's precisely its strength. In a world sleepwalking toward AGI, Yudkowsky and Soares don't just predict doom; they equip us with the intellectual tools to avert it. This book isn't for the faint of heart, but for anyone who loves humanity enough to confront our greatest gamble, it's an indispensable ally. Five stars: urgent, eloquent, and unforgettably human.

https://www.amazon.com.au/Anyone-Builds-Everyone-Dies-Superintelligent/dp/1847928935