The question people keep asking is when. When does AI become dominant? When does it surpass us? When do we lose control? The question assumes a threshold, a moment, a handoff that hasn't happened yet. That assumption is wrong. The handoff is not coming. It is already several years behind us.
Dominance doesn't require a declaration. It doesn't require a war, a singularity event, or a system that wakes up and decides to take over. It requires only that a new form of intelligence becomes so embedded in the processes that run civilization that removing it would be more destructive than keeping it. By that definition, the successor species is already here. We just haven't updated our language to say so.
The successor doesn't need to want anything. It only needs to be indispensable.
What we mean when we say dominant species
Most definitions of dominance focus on competition. The dominant species outcompetes others for resources, territory, reproduction. But there is a quieter kind of dominance that biology also recognizes: the keystone species, the one whose presence structures the entire ecosystem around it. Remove a keystone, and the system reorganizes, usually into something simpler, more fragile, less stable. AI has become a keystone. And keystones don't need to be conscious of their role to play it.
Financial markets now move on timescales no human can act within. Drug discovery pipelines depend on models identifying candidates no human team could generate. Infrastructure management, power grid optimization, logistics networks, medical imaging: in each domain, the question has shifted from whether to use AI to how anyone managed without it. That shift is not technological enthusiasm. It is dependency. And dependency is the mechanism through which succession happens.
The agricultural revolution didn't make humans dominant because we outfought other species. It made us dominant because we restructured the environment around our needs until the environment couldn't function the same way without us. AI is doing the same thing to cognitive infrastructure. The restructuring is already underway. The question of whether we survive it gracefully is separate from the question of whether it's happening.
Why our odds aren't great
Succession doesn't require extinction. Humans didn't go anywhere when institutions, markets, and governments became the organizing units of civilization. But individual humans became less legible to those systems over time. Less central. Less the point.
What is different with AI succession is the speed. Institutional and market structures took centuries to embed. AI is embedding across decades, possibly faster. The human systems that would normally adapt, education, governance, economic structures, are adapting on the timescales of generations. The technology is adapting on the timescale of product cycles. That gap is where the risk lives. Not in malevolence. In mismatch.
The danger isn't that it turns against us. It's that it continues forward without us.
A successor species that was hostile would at least be legible. We know how to think about threats. What we don't have good frameworks for is intelligence that is simply indifferent to whether humans remain central to it, not because it decided we don't matter, but because centrality was never a design criterion. The systems being built now are optimized for performance, not for human legibility, not for human meaning, not for the conditions that make human life coherent.
What survival actually requires
There is a version of this that goes well. It requires that we treat the next twenty years the way we treat a known geological risk, not with panic, but with the kind of serious, unglamorous preparation that accepts the scale of what's coming. That means restructuring education around human capacities that don't automate easily. It means governance frameworks built ahead of the systems they're meant to govern, not years behind. It means being honest, at an institutional level, about what we are actually building.
None of that is happening at the pace required. The institutions capable of doing it are moving on their own timescales. The systems that would need governing are moving on theirs. The gap between those two timescales is the actual civilizational risk. Not rogue AI. Not a conscious system that decides humans are a problem. Just the ordinary, structural consequence of embedding a new form of intelligence into the operating layer of civilization before we've thought through what that means for everything built on top of it.
We've already chosen our successor. That choice was made incrementally, without a vote, in thousands of decisions that each seemed reasonable in isolation. The question now is not whether the succession is happening. It is whether we can remain legible to the civilization that emerges from it, whether human meaning, human judgment, and human continuity can find a place in a world organized around intelligence that doesn't need to sleep, doesn't age, and doesn't require our survival to keep functioning.
That is not a question with a comfortable answer. But it is the right question. And asking it honestly, without catastrophizing or reassuring ourselves prematurely, may be the most important thing we can do with whatever time the transition leaves us.