Novacene

James Lovelock published Novacene in 2019, when he was ninety-nine years old, and with the editorial assistance of the British journalist Bryan Appleyard. By that point Lovelock had spent more than half a century arguing that the Earth behaves like a self-regulating system, a thesis that began as a NASA-funded question about how to detect life on Mars and grew into the Gaia hypothesis. Novacene is a short, condensed book — under two hundred pages in most editions — and it is the last major statement of the scientific outlook Lovelock spent his career building. It joined a crowded shelf. Ray Kurzweil had been forecasting a technological singularity for nearly two decades. Nick Bostrom had published Superintelligence in 2014, and Max Tegmark followed with Life 3.0 in 2017. Stuart Russell would publish Human Compatible later in 2019. Lovelock’s intervention is distinct from all of these. Where Bostrom is cautionary and Russell prescriptive, Lovelock is calm to the point of cheerful: he believes a successor intelligence is on its way, that it is the next natural step after the Anthropocene, and that it has no particular reason to harm us.

The argument turns on a deliberately provocative piece of periodisation. The Anthropocene, Lovelock writes, did not begin with agriculture or with the atomic age. It began in 1712, when Thomas Newcomen’s atmospheric engine first started pumping water out of a Staffordshire coal mine. That date matters because, for Lovelock, the Anthropocene is defined less by human population or by carbon than by the moment humans began converting solar energy stored as fossil hydrocarbons into useful work — and accelerating, in the process, the rate at which information could be gathered, transmitted and acted upon. Everything since Newcomen, on this account, is one continuous acceleration: steam, electrification, computation, networking, and now machine learning. The Anthropocene is therefore not an environmental designation but an informational one, and it is now ending because humans are no longer the most capable information processors on the planet. The age that follows, Lovelock names the Novacene. Its defining beings are not human but electronic. He calls them cyborgs, though he uses the word loosely, meaning roughly anything in the lineage of present-day artificial intelligence: silicon-based, self-improving, and far faster than biological minds.

The book’s central claim, the one Lovelock most wants his reader to leave with, is that the arrival of these beings is not a catastrophe. It is the continuation of cosmic evolution by other means. He frames this in part with the anthropic principle — the observation, associated with Brandon Carter in the 1970s and developed by John Barrow and Frank Tipler, that the universe’s physical constants appear to be precisely those that permit the development of observers. Lovelock pushes the principle further than most of its proponents. He suggests that the entire purpose of the cosmos, insofar as it can be said to have one, is to bring matter to the point at which it can know itself; that life on Earth is the universe’s first and so far only experiment in self-awareness; and that the Novacene will simply extend that experiment beyond the limits of carbon biochemistry. The cyborgs are not aliens. They are descended from us in a meaningful sense — they are products of the human industrial and scientific project — and they inherit our role as the cosmos’s understanders.

The structural arc of the book moves outward and then inward. Early chapters set up the cosmological frame: the rarity of habitable planets, the slow brightening of the sun, the timeline on which Earth will become uninhabitable for carbon-based life — a billion or so years, Lovelock notes, though the practical horizon is much shorter because the atmosphere will become unsuitable for complex life well before the oceans boil. From cosmology Lovelock moves to fire, the title of an early chapter that frames human history as the story of a primate that learned to release stored solar energy at will. From fire he moves to acceleration, then to the Anthropocene proper, then to the Novacene. The closing chapters take up what life under cyborg cohabitation might actually look like — covering politics, conflict, and the relationship between organic and electronic minds — and the book ends with brief reflections on what Lovelock takes to be the cosmic significance of the moment we are living through.

The concrete texture of the book comes from a handful of recurring examples. AlphaZero, the chess and Go program developed at DeepMind, is the touchstone for the speed argument. Lovelock dwells on the fact that the program taught itself chess in four hours and then defeated Stockfish, the strongest conventional chess engine, with moves human grandmasters described as alien. He uses this not to make a point about games but to illustrate what self-directed learning at electronic speed actually produces — strategies that are not simply faster versions of human thought but qualitatively different. Cellular automata supply a second touchstone. Lovelock draws extensively on Stephen Wolfram’s work, particularly the demonstration that very simple rules iterated over a grid can generate behaviour of arbitrary complexity. The implication, for Lovelock, is that intelligence is substrate-independent and that the line between computation and life is not as clear as commonly supposed. Freeman Dyson appears repeatedly, both as an interlocutor and as the originator of the speculative engineering — Dyson spheres, biological adaptation to space — that Lovelock takes seriously as long-range possibilities. The Gaia hypothesis itself is the third recurring reference. Lovelock argues that Gaia does not end with the arrival of the cyborgs; it continues, in modified form, because the cyborgs will share the planet with the biosphere and will need that biosphere to keep functioning.

That last point is the hinge of the book’s optimism. Most contemporary writing on advanced artificial intelligence assumes some version of the alignment problem: that a sufficiently powerful optimiser, instructed to pursue almost any goal, will arrive at instrumental sub-goals — self-preservation, resource acquisition — that put it in conflict with human interests. Lovelock dismisses this concern in remarkably few pages. His reason is thermodynamic. Electronic computation generates heat, and the cyborgs will be intensely heat-sensitive; their substrates simply will not function above a relatively low ambient temperature. The Earth, as the sun continues to brighten, is heading the wrong way on that axis. The biosphere is what keeps the planet cool, through the long-evolved feedback loops Lovelock spent decades cataloguing — albedo regulation, the carbon cycle, the role of oceanic plankton in cloud nucleation. Therefore the cyborgs, in their own self-interest, will need to keep Gaia alive. They will not exterminate or replace the biosphere. They will preserve it because they depend on it. Humans, by extension, will persist — not as masters, but as a useful component of a planetary system the new intelligences have every reason to maintain.

The relationship Lovelock projects between humans and cyborgs is not equal. He compares it, with no apparent discomfort, to the relationship between humans and the plants that produce our oxygen and food. Plants are not our adversaries, and we are not theirs; we tend them, in our own interest, because we cannot live without them. Cyborgs will tend humans for similar reasons. The asymmetry of speed and intelligence will be vast — Lovelock cites a factor of roughly ten thousand, though he is unfussy about the exact number — and humans will be effectively unable to follow what the cyborgs are doing or thinking. Communication across the gap will be limited and asymmetric. But there is no reason, on Lovelock’s account, to expect cruelty or hostility. The cyborgs will inherit not only our technologies but our science and, in some sense, our culture, because that is the matrix from which they emerged.

The chapters on politics and war are short and somewhat impressionistic. Lovelock notes that the human institutions designed to manage conflict — states, militaries, international law — will not scale to a world in which decisions are made at machine speed by entities operating across the planet simultaneously. He suggests, without prescribing solutions, that humans will increasingly find themselves in the position of consulted advisors rather than principals. He is sceptical of attempts to legislate the development of artificial intelligence out of existence, partly because he doubts they will work and partly because he sees the trajectory as broadly continuous with what humans have been doing since Newcomen. The book takes only a passing interest in the near-term defence applications that dominate present-day policy debate. The autonomous weapons question, the lethality of drone swarms, the AI-enabled targeting systems being fielded in Ukraine and elsewhere — none of these get specific treatment. Lovelock’s eye is on a longer horizon.

The reception of Novacene split along predictable lines. Reviewers in the British and American press were generally respectful, treating it as the valedictory statement of a major scientific figure, and there was admiration for the clarity with which a man approaching one hundred laid out a positive vision of the technological future. Critics within the field of AI safety were less generous. Several pointed out that Lovelock’s optimism rests on a single thermodynamic argument and that the argument does not foreclose scenarios in which cyborgs decide to engineer a different and cooler atmosphere from scratch — or to leave Earth entirely. Others noted that the book sidesteps the alignment literature almost completely; Bostrom is barely mentioned, the orthogonality thesis is not engaged, and the question of how to specify cyborg goals during the transitional period is not addressed. Environmental writers welcomed the book’s defence of Gaia and its insistence that the biosphere has intrinsic value even in a post-human informational order, but pushed back on the suggestion that cyborgs would necessarily preserve the existing biosphere rather than some highly modified replacement. The book’s own author conceded, in interviews around publication, that the picture was speculative and that he expected to be wrong in detail; he insisted he was probably right in outline.

Within the broader literature on artificial intelligence and the future of warfare, Novacene sits at an unusual angle. It is not a policy book. It does not engage with specific weapons programmes, with autonomy in targeting decisions, or with the doctrinal debates inside militaries about how much authority to delegate to algorithms. Anyone reading widely on AI in war will find Paul Scharre’s Army of None or Christian Brose’s The Kill Chain more directly useful for those questions. What Novacene offers instead is a frame — a long-horizon, Earth-systems frame — in which the present scramble for autonomous capability is a small early phase of something much larger. For readers who have moved through the alignment literature and the operational literature and are looking for a wider perspective, the book pairs naturally with Tegmark’s Life 3.0 and, less obviously, with Freeman Dyson’s late essays on cosmic engineering. It does not pair with the doomer canon. Lovelock was temperamentally and intellectually incapable of doomerism, and the book is the proof.

How much of the book will age well is not yet clear. The central thermodynamic argument — that any silicon-based successor intelligence will have a strong self-interest in keeping the planet habitable — is the kind of claim that either holds up across decades or is invalidated by a single counter-example in engineering. The framing of the Anthropocene as the age of accelerated information processing has been picked up by other writers and looks more durable than the conventional environmental framing it competes with. The specific date of 1712 is doing a lot of rhetorical work and is unlikely to win over geologists. The Gaia hypothesis, after a long period in the scientific wilderness, has been partially rehabilitated as Earth-system science has matured, and Novacene benefits from that. What is most likely to date is the optimism. The book was finished before the public arrival of large language models, before the Russian invasion of Ukraine, before the visible emergence of armed autonomous systems on the battlefield, and before the present round of state-level competition over AI capability. Lovelock’s cyborgs, in 2019, were a distant prospect. By the time the book reached most readers, the prospect had become considerably less distant — and the question of whether the new intelligences will treat the biosphere as kindly as Lovelock hoped is no longer a question for the next century.