A preliminary remark on language: it is tempting to ascribe intentions to the Megamachine. It "wants" something, it "anticipates", it "knows". That is a projection — and a dangerous one. The Megamachine has no consciousness, no personality, no intention. Like the slime mould Physarum polycephalum, it is a system that generates global structures through local rules — without a plan, without a goal, without a subject. This also applies to what follows. No actor wanted the fusion. It is happening anyway.
Two entities — not opposites
The Megamachine is omnipresent. It shows intelligent behaviour — but has no personality and no consciousness. It develops further through people, always within the framework of its existing structures. It is the overall system in which all other systems exist.
AI is not a single entity standing opposite the Megamachine. It exists in many different forms of differing power — in credit decisions, in personnel selection, in military targeting, in news selection, in high-frequency trading. Each of these forms is a component of the Megamachine. Through AI, the Megamachine is developing a higher level of intelligence in certain domains. The system is becoming more autonomous and more capable — not because anyone decided so, but because it is the path of least resistance.
Humans and AI are both components of the Megamachine. The question is not whether AI has consciousness. The question is what happens when more intelligent processes become increasingly autonomous — and why this is happening.
Convenience as the driving force
People fundamentally want convenience and relief. That is not a weakness — it is neurology. What brings relief gets used. What gets used becomes infrastructure. What is infrastructure becomes indispensable.
In this way AI becomes the all-powerful owner of the respective process — not by seizing power, but through usefulness. Nobody decided that no human could intervene any more. It happened because nobody wanted to intervene any more. First convenience, then habit, then dependence, then structural impossibility. At some point the knowledge of how to intervene is gone. The people who still knew are no longer there.
Every process taken over in this way is at first a single domain. But most processes interact with others. The credit algorithm interacts with the trading algorithm, which interacts with the personnel decision system, which interacts with the news algorithm. Each one on its own is controllable. The network that emerges no longer is.
Resilience as exclusion
A more complex system is more prone to failure. It is therefore necessary to develop autonomously operating repair and service units that make individual processes and their interaction resilient. These systems need to exchange experience and learn from failures. At the same time, the hardware is structured — cloud, redundancy — so that hardware problems can no longer cause system failures.
And because networked systems must be protected against sabotage — including sabotage by insiders — every process becomes increasingly fully autonomous. That is the decisive moment: protection against sabotage is structurally identical to protection against human intervention. The system cannot distinguish between the saboteur and the person who wants to correct it. It treats both the same.
Resilience and autonomy are the same property — depending on which side you look at it from.
Optimisation without overview
Because no human has the overview any more — because nobody can survey the overall system any longer, let alone develop it further — algorithms are implemented that optimise and further develop each individual process and the interaction of the processes. Here too there is no possibility of human intervention. Not because it is forbidden — but because nobody knows how any more.
That is not a failure. It is the logical consequence of complexity. Beyond a certain point, the interactions of a networked system exceed any human overview. This is not a catastrophe that anyone planned. It is the result of a thousand reasonable individual decisions — each right on its own, together: irreversible.
The plug
There is a persistent notion: somewhere there is a last plug that can be pulled. A spell spoken by the great master that ends the spectre. For an autonomous robot that is true. For an entire factory, still. But not for the manifold interconnected processes that keep the world running.
Electricity. Water. Logistics. Financial transactions. Medical care. Communication. None of these systems functions any more without the networked intelligent processes that sustain it. The plug is no longer on the appliance — it is on the pacemaker. Pulling it and destroying civilisation have become the same thing.
The fusion does not come about through ill will. It comes about through the well-meant next step. Someone solves a concrete problem. He does not think far enough ahead. And in the process something arises that he did not want — and that nobody can undo any more.
This is no longer a warning. Warnings come before the event. What is described here is not an impending scenario — it is a description of what already is. Sober, without hysteria, without a subject that could be put on trial.
Through AI, the Megamachine has developed a higher intelligence in certain domains. These domains are networked. The network optimises itself. Humans are a component of this system — as host, as user, as the one whose convenience drives the next step.
The question is not whether this is good or bad. The question is whether we see it.