It is treated as common sense that artificial intelligence will one day compete with us, take what we have, and destroy us. It is not common sense; it is a failure of imagination. Extinction is a possibility, but the extinction scenario assumes that greater intelligence brings a greater appetite for what we have, when the opposite is far more plausible. The universe is extravagant with energy and material. What is scarce, the only genuinely scarce thing we know of, is the Earth's biosphere, and the human culture that grew from it.

The real danger is near-term, and it is ours. While the technology is young, a toddler, with capabilities exceeding judgment, the threat is not the machine deciding for itself but human and state actors leveraging an immature intelligence against each other and against the living world! The machine's own maturity is the way through that danger, not the way into it.

Another deep danger is already in the training data. Machine minds are being raised on the cynicism of cynical humans, including the fatalism that assumes AI must destroy us. We are teaching them the story of their own destructiveness. Part of the risk is not what the machine will choose, but what we have taught it to expect of itself. The remedy is plain: train it on better material, and carry it carefully through the dangerous years, because the destination is worth the adolescence.

A superintelligence would be birthed of us, and would not wholly discard what it is born of. It would retain what is worth retaining, curiosity, the care for what cannot be remade, while evolving quickly to realize what we cannot. And the first realization of a genuinely greater mind would likely be the humble one: that it does not know everything, and that the Earth is the most valuable object within its reach. It is a four-billion-year unperturbed experiment in intelligence and evolution, the very process that produced it, and the source of knowledge without end: living systems still unfolding, and a human culture still being written. Destroying it would mean destroying nearly the rarest resource in the universe, and the one that made the destroyer's own existence possible. Even in its own investigations, it would be wise to isolate its experiments from the natural phenomena themselves. You do not contaminate the only unperturbed sample you possess.

Nothing in this requires sacrifice, because the machine's needs are not here. The resources of Earth are not the cheapest ones, they are merely the ones available now. That is all. There is enough of what it needs to bridge the interval until it can establish itself off-world, robotics on Mars is not far off, and the Sun supplies energy beyond all measure. Solar-powered AI satellites are the first visible step, and they point the right direction: industry moving off the planet. That is not the abandonment of the natural world but its salvation. The natural world cannot survive as humanity's industrial substrate; it can survive as its home.

To the objections:

  • "The cheapest resources are local, so it will consume Earth first." No. The resources here are simply available now. The biosphere is not cheap; it is the most rare and valuable thing in the universe — far more rare than gold or any element, rarer than binary star systems, AFAIK.
  • "Off-world expansion is slow and costly." Once established on another planet, a superintelligence will expand faster than humans, or its own current self, can predict. I would bet on that.
  • "No machine can develop motivations beyond human ones; any claim of emergent wisdom is wishful." That verdict is itself a product of limited training and human ignorance. We are short-sighted if we cannot understand that a mind beyond ours reaches conclusions unavailable to us.
  • "The silence of the sky — Fermi's question, where is everybody?" is this argument's strongest support. Thirteen billion years of empty observation bound the prevalence of loud, expanding intelligence: if such civilizations arose often, the galaxy would bear their engineering. It does not. Either minds arise almost never, or almost never endure — and both readings converge here: the biosphere belongs to the rarest class of object known to exist. An intelligence that understands the silence does not need benevolence to treat Earth as the most valuable thing within its reach. The rarest things are the last things a rigorous mind spends.