Machines of Loving Grace: The Quest for Common Ground Between Humans and Robots
Source: https://www.harpercollins.com/products/machines-of-loving-grace-john-markoff ↗
Markoff, who covered technology for the New York Times for three decades, structures the history of computing around a fundamental tension: artificial intelligence (replacing human capabilities) versus intelligence augmentation (extending them).
He traces both traditions from their shared origins at Stanford and MIT in the 1960s — McCarthy's AI lab on one side, Engelbart's augmentation research on the other — through robotics, autonomous vehicles, and personal computing to the present.
The institutional history is meticulous: who funded what, which labs competed, how DARPA's priorities shaped both fields.
Markoff's central argument is that the choice between automation and augmentation is not technical but ethical and political, and that Silicon Valley has oscillated between the two without resolving the tension.
Written just before the deep learning explosion, the book provides the historical framework needed to understand why the AI-versus-augmentation debate keeps recurring.
Central argument
Markoff argues that the history of computing has been shaped by a persistent, unresolved tension between two competing visions: artificial intelligence, which seeks to replace human capabilities, and intelligence augmentation, which seeks to extend them. These traditions emerged from rival research cultures in the 1960s — McCarthy's AI lab versus Engelbart's augmentation program — and were shaped less by technical inevitability than by funding priorities, institutional politics, and ethical choices. His central thesis is that the oscillation between automation and augmentation in Silicon Valley is not a technical problem awaiting a solution, but a recurring political and ethical choice that the industry has consistently avoided making explicitly.
Critique
The book's framing as a binary tension between AI and augmentation, while analytically clarifying, risks imposing a false dichotomy on a landscape where the two traditions increasingly interpenetrate — large language models, for instance, are simultaneously automating tasks and augmenting cognition in ways that resist clean categorization. More significantly, published in 2015 and therefore predating the deep learning inflection point, the institutional and political dynamics Markoff describes may not fully account for how the concentration of compute and data in a handful of companies has restructured the terms of the debate in ways that make the 'choice' between automation and augmentation far less available to most actors than his ethical framing implies.
Why it matters for product
For a CPO, the AI-versus-augmentation axis is not an abstract philosophy — it directly governs product strategy decisions about where AI gets embedded: replacing user actions entirely (automation) versus surfacing information and expanding user agency (augmentation). Teams that have not made this choice explicitly tend to make it implicitly and inconsistently, producing products that neither automate reliably nor augment meaningfully. Markoff's institutional history also serves as a diagnostic: the reason product organizations keep revisiting this question without resolution is structural, not analytical, which means the answer requires a deliberate organizational stance, not just a feature roadmap.