The framing device is in the title. Life 1.0 is biology that cannot change its hardware or its software within a lifetime, meaning simple organisms that evolve and do not learn. Life 2.0 is us, stuck with our hardware but able to rewrite our software through learning and culture. Life 3.0 would be able to redesign both.
It is a neat device, it carries the argument well through the first chapters, and it is the sort of framing a physicist reaches for because it makes a messy question look tractable. That is both the strength and the recurring weakness of the book. Take the strengths first. This is far and away the most readable book in its category.
The prose is warm and direct, the author clearly enjoys explaining things, and the structure moves at a good pace. Compared to Bostrom, which is the obvious comparison and which reads like a taxonomy, this one wants you to keep turning pages and mostly succeeds. The book opens with a fictional prologue about a team that develops a powerful system and uses it to accumulate money and influence. As fiction it is unremarkable.
As a device for making an abstract worry concrete it works well, and it gets you to think about the mechanics of how capability turns into power, which is the interesting question and is more useful than any amount of abstract argument about intelligence explosions. The near term chapters are the best material in the book. Autonomous weapons, the labour market, legal responsibility when a machine causes harm, and the concentration of economic power. These are subjects where he can be concrete because the developments are already visible, and the writing is at its most useful when it stays there.
The chapter on autonomous weapons in particular is measured and specific, and it stands up better in 2026 than most of what was written about the subject at the time. He also declines to prescribe. The middle of the book lays out a set of possible long term outcomes, from benevolent stewardship to various dystopias to human extinction to scenarios where humanity chooses to stop, and rather than nominating a winner he asks the reader to notice which ones they find appealing and why. That is a genuinely good exercise and more intellectually honest than a book that arrives with the answer.
Now the problems. The physics framing overstays. There is a long stretch about the ultimate physical limits of computation, the energy that could theoretically be extracted from matter, and how far a civilisation could spread through the accessible universe. It is interesting in the way a popular cosmology book is interesting and it is doing no work in the argument.
Speculation dominates the second half. Chapters on consciousness, on goals in physical systems, and on what a very distant future might contain, are written in the same confident, engaging voice as the chapters on autonomous weapons, and there is no signal to the reader that the epistemic status has changed completely. A reader without background will not notice the transition from things we can observe to things nobody can know. Technically the book is thin.
You will not learn how any machine learning system works. There is no real engagement with the mechanics, only with capabilities treated as a black box that gets better. For a book about the implications of a technology, that is a defensible choice and it means the reasoning floats free of any constraint from how the technology actually behaves. And the 2017 vintage is visible.
The examples are of their moment, the sense of what was close and what was distant has not aged well in either direction, and the specific worries emphasised are not the ones that turned out to dominate the following decade. The general shape holds up better than the particulars. My three point five is for an enjoyable, humane and well structured introduction that gets a lot of people thinking about the right questions, marked down for speculation that runs far past the evidence in the same confident tone, for a physics frame that outlives its usefulness, and for a complete absence of technical grounding. Read it first, then read something with more rigour, and notice how differently the two feel.