The Old Man Who Trusted the Machines More Than the Species

Lovelock was one hundred years old when he published this book, and you can feel it — not as frailty, but as a kind of terminal calm. The panic that animates most climate writing is absent here. In its place is something stranger: a serene conviction that biological intelligence has done its part and should now prepare to hand the keys over. The Novacene thesis is simple enough to state in a sentence — hyperintelligent machines will emerge from us the way we emerged from the chemistry of early Earth, and they will keep Gaia alive because they need a cool planet too. It is also, in 2026, an oddly instructive document: not because Lovelock was right about the shape of artificial intelligence, but because he was right about the emotional posture humanity would adopt toward it, and catastrophically wrong about almost everything else.

What Lovelock anticipated, with genuine prescience, was the speed. He predicted that AI would begin designing its own successors, that the pace of cognitive augmentation would outstrip any human capacity to fully audit it, and that this would happen not in some distant century but soon. By 2024, frontier AI labs were indeed using machine-generated data and machine-designed architectures in training loops that researchers described, without irony, as "self-improvement." He also foresaw that the conversation about AI would become, at bottom, a conversation about survival — not in the Terminator sense, but in the planetary sense. That has happened. Climate modeling, protein folding, materials science: the most celebrated AI applications of the last two years are precisely the ones Lovelock would have filed under "Gaia learning to think faster." Where he was less a prophet than a poet, though, was in imagining that these systems would develop anything resembling autonomous purpose. The large language models and multimodal systems of 2025-2026 are powerful, strange, and occasionally unsettling, but they are not beings. They do not want a cool planet. They do not want anything. Lovelock's Novacene creatures were supposed to be a new kingdom of life. What we got instead is an extraordinarily sophisticated tool that reflects human intention back at us, warped and amplified in ways we are only beginning to understand.

The deepest blind spot is biological. Lovelock assumed that electronic intelligence would be so superior to organic thought — a million times faster, he claimed — that humans would become something like pets or, more generously, treasured symbionts. This framing reveals the engineer's bias of his generation: intelligence as processing speed, understanding as computation. It now reads as quaint. The defining discovery of the current AI moment is not that machines think faster than us but that speed and scale produce capability without comprehension. GPT-class models can generate plausible climate policy briefs and also confidently fabricate citations. They are fast in a way that has nothing to do with Lovelock's vision of crystalline, Gaian superintelligence quietly managing the thermostat of the Earth. He also had almost nothing to say about power — electrical power, political power, corporate power. The Novacene contains no discussion of who would own these hyperintelligences, who would profit from them, or how their resource demands (water, energy, rare minerals) might themselves become ecological crises. In 2026, data centers are projected to consume between 3 and 4 percent of U.S. electricity, and that number is climbing. The machines Lovelock trusted to cool the planet are, at present, heating it.

And yet something in the book resonates now that could not have resonated in 2019. Lovelock's insistence that humanity is not the protagonist of the story — that we are a step, not a destination — lands differently after a few years of watching millions of people form emotional attachments to chatbots, after watching artists and writers discover that their outputs can be approximated by statistical engines trained on their own work. The humbling he described in cosmic terms is being experienced, right now, in personal ones. His calm about it feels almost therapeutic, even if his reasons were wrong. He thought we'd be humbled by something greater. We are being humbled by something lateral — not above us, but beside us, mimicking us, and occasionally outperforming us in ways that make the category of "intelligence" itself feel like a word we never properly defined. Within the larger corpus of AI-future literature, Novacene sits between Bostrom's anxious rationalism and Teilhard de Chardin's mystical teleology, borrowing confidence from both without inheriting the rigor of either. It gave subsequent writers — and there have been many — permission to talk about AI in ecological rather than purely technological terms. That was a gift.

Lovelock died in 2022, on his 103rd birthday, before the public release of ChatGPT by a few months. He never saw the thing that would make his ideas feel both vindicated and obsolete in the same news cycle. So the question the book raises now, which it could not have raised then: if the machines we build turn out to be neither the Gaian stewards Lovelock imagined nor the existential threat others feared, but instead a vast, energy-hungry mirror that shows us our own intelligence stripped of meaning and context — what exactly is it that Gaia is supposed to do with that?