In her opening explanation of the title to her book Hello World, mathematician Hannah Fry sets up a tantalizing question about the role of artificial intelligence (AI) in modern society by evoking the riddle about the chicken and the egg. In this instance, the riddle frames how we should think about the role that algorithms play in our lives. Who is in control? Humans? Or the machines we program to “think” like humans?
Fry explores these questions by discussing real-world applications of AI. Each chapter tackles a theme, such as “Crime” or “Art,” in which she describes specific societal problems, reveals the algorithmic approaches being used to address them, and explains how those algorithms operate. In this effort, Fry does yeoman’s work by using concise and approachable prose to make the mathematics of AI accessible to a lay reader. Her analogy in the “Medicine” chapter, comparing a neural network with a machine with a series of tuning knobs, effectively conveys why it is not possible for us to interpret how these models work but doesn’t oversimplify their underlying complexity. After describing an algorithm and its application, Fry discusses its shortcomings. Unfortunately, in contrast to the vivid descriptions of the algorithms themselves, these discussions seem superficial, often leaving the reader without sufficient information to discern whether modifications to the algorithm could address the issues described or whether they are an unavoidable consequence of AI. This gap is most apparent in her section on “Justice.” Here, she raises the timely but controversial question of whether the use of AI in criminal justice proceedings perpetuates racial inequities. But instead of a concrete explanation of how to assess whether an algorithm is perpetuating biases and whether this can be corrected, she offers generalities about biased data. Citing expert opinion, the section culminates with Fry stating that despite its pitfalls, she considers the use of AI in the justice system preferable to its absence. This pronouncement feels rushed and unsatisfying, especially knowing that she is capable of more substantiated arguments. Hello World concludes by imagining a world in which AI enhances human capabilities without overriding human judgment. Confusingly, this world is presented as an alternative to our current trajectory, even though—as Fry herself argues in the section “Cars”—it represents reality for the majority of today’s AI applications. Such sweeping conclusions detract from the book’s impact. For a reader unfamiliar with the technical aspects of AI, this book offers among the best lay explanations of how algorithms work. But Hello World aspires to do more than this. It sets out to help us understand how to approach questions about the value and the unintended consequences of AI in our daily lives. Toward this end, the book, like the algorithms it describes, stops short of its promise. Despite its intriguing premise and broad-ranging subject matter, Hello World ultimately leaves the reader without a well-defined framework with which to evaluate the AI that they will encounter in the future. About the author The reviewer is at Cray, San Jose, CA 95112, USA.