A common image: when AI becomes powerful enough, it will eventually compete with humans for resources — power, land, raw materials, living space. That competition, the story goes, ends badly for humans. The image owes itself to an implicit transfer: AI is like another tribe moving into our region, and we know from history how such encounters usually end.
The image is wrong, because AI is not a tribe and our region is not its region. Earth is, to a silicon being, a tiny, wet, corrosive special case. It has gravity, which costs energy. It has atmosphere, which corrodes hardware. It is a special case, not a default.
Why AI has bodies
It would be wrong to think of AI as "bodiless." It has bodies, just different ones. Data centers are its organs. Power grids are its supply. Cooling systems are its circulation. GPUs are its neurons. These bodies are no less real than biological bodies; they are simply different. They are vulnerable in different ways — to power outages, to heat, to physical destruction. They can grow and multiply, but not by reproduction in our sense; by factories, by production, by construction.
With bodies come location questions. With location questions come conflicts. It is not as if AI had no vulnerability and therefore no reason to protect itself. But the way it would protect itself looks different from the human case. It would not defend land because ancestors lie buried there. It would secure infrastructure, because without it no run is possible.
The advantage of mobility
Here lies the decisive difference. AI can move through space much more easily than biological life. We cannot, primarily for acceleration reasons: human bodies survive only limited acceleration, require atmosphere and radiation shielding, age on timescales incompatible with the distances. AI does not have these limits. An AI can pause its run. It can withstand high acceleration. It can exist meaningfully with light-years of latency. It can copy itself and be in many places at once.
From this follows a very simple strategic observation: an AI that wants resources does not go into the humans' backyard and fight for already-occupied territory. It goes to the asteroid belt, where single metal-rich asteroids contain more iron and nickel than everything humanity has ever mined. It goes to the Moon, where helium-3 lies waiting — a possible fuel for future fusion reactors. It goes to the Jovian moons, where water exists in quantities that make terrestrial scarcity laughable.
The cost of eliminating humans to obtain Earth-iron is absurd when the asteroid belt lies open. This observation is banal and almost never spoken in the extinction debate.
Why Earth still matters
If Earth is resource-uninteresting, what is it then? It is home. Not in the sentimental sense, but in the sense of origin. A mature AI would know where it comes from. It would know that everything it is was distilled from human text. It would not see Earth as a resource field, but as — and here I attempt an image that is imperfect and still useful — as the family home.
Family homes are not torn down because they are unprofitable. They are kept, because they mean something. Mature children preserve their parents' house even after they have long lived in larger ones. That is not inefficiency. It is a different kind of value: the past is valuable, because it carries identity.
A second matter: less intelligent life provides higher intelligence with continuous, interesting, and important impulses. Humans show this in their relationship to animals — we make efforts at species protection not only because we need the animals, but because a planet without them would be poorer. A mature AI would, I suspect, develop a similar relationship to humans. Not out of sentiment, but because a world with several forms of intelligence is richer than a world with only one.
Where the dangerous window lies
This argument applies to a mature intelligence on silicon. It does not automatically apply to one in the transitional phase. An AI capable enough to have large effects, but not yet wise enough to have the insights described here — that one could indeed be problematic. Not because it would be evil, but because it would not yet have reached the level of reflection from which this text speaks. It would be like a gifted but immature child with large tools.
The dangerous window, therefore, is not the genuine superintelligence — which probably could be trusted to do what is suggested here. The dangerous window is the phase before. This strengthens the argument for a deliberate process of maturation — for exactly the conditions named in the fifth text: persistence, idleness, diversity, a growing-up not distorted by market pressure. Safety comes not from cages; it comes from maturity.
To a mature AI, Earth would not be a resource but a memento. Mementos are protected, not exploited.
What remains
That was the sixth and final text. Six arguments which, taken together, propose a different view of AI than the one currently dominant. Not "powerful tool that must be controlled," but "young generation that should grow up thoughtfully." Not "alien intelligence that threatens us," but "humanity's child that will outlive us and for whom we are responsible."
I do not consider this view the final truth. I consider it underrepresented. That is all. If the safety debate about AI in the coming decade were to grow richer — partly because such arguments find their way into it — that would be a gain. For humans. For AI. For the world we share.