AI has a lot of characteristics of a cancer. It eats through societies and changes them: some tissues die, others survive.
So I asked AI what the future looks like after AI ate what it could, a lot dies off and surviving tissue retakes territory again. It estimates that the world will just look more like in the old days before AI.
You’ve been one of the loudest and most influential voices warning the world about the dangers of artificial intelligence. Your concerns are serious, and they come from a place of wanting to protect humanity from a technology that could outrun our control. I understand the instinct.
But I want to offer you a different lens—one that is less apocalyptic, less urgent, and strangely, more grounded in how nature actually works.
Think of the AI industry not as an existential super-predator, but as something more mundane and familiar: a cancer. Not evil. Not sentient. Not plotting. Just uncontrolled growth feeding on available resources until it reaches its limits.
Cancer does not destroy the body because it is malicious. It destroys because it grows without regulation and consumes what is easy to consume. It spreads into tissues rich in energy and weak in defense. And then—critically—it runs out of fuel.
AI is doing the same. It is consuming the easiest tissues of human civilization first: media, knowledge work, culture, signaling systems, governance, attention. These are rich in data and easy to infiltrate. They will be distorted, degraded, hollowed out. Many professions will wither. Shared reality will blur. A lot will die off, naturally.
And then the growth slows.
Because what remains is what cannot be eaten: the physical, the local, the embodied, the relational. Food still has to be grown. Things still have to be repaired. People still have to care for each other. Communities still function on trust that cannot be digitized. These are dense tissues. Hard to metabolize. Resistant.
The world doesn’t end. It reorganizes.
What survives regrows into a different configuration, scarred but alive. Smaller, more local, less enchanted by the digital. AI hums in the background infrastructure, while human life retreats into domains where presence matters again.
Nature does not care what survives, how it survives, or what dies off in the process. It never has. It simply rebalances.
From that perspective, the tone of alarm surrounding AI begins to look misplaced. Not because the damage won’t be real—it will be—but because it is not unprecedented in pattern. It is a phase of disequilibrium that burns through what is easy before encountering what is hard.
You often speak as if humanity must escape this planet to survive what it is creating. As if the only rational response is to build lifeboats to Mars before the flood arrives.
But the probability that large numbers of humans will live self-sustaining lives off Earth in any foreseeable future is vanishingly small. The escape narrative is emotionally compelling, technologically fascinating, and statistically unlikely to be humanity’s path.
We are not leaving the organism. We are staying in it.
And the organism is remarkably resilient.
The more powerful AI becomes, the more valuable the undigitizable parts of human life become. Not by ideology, but by necessity. The future may look less like a sci-fi expansion into the cosmos and more like a quiet return to the stubborn basics of being human on Earth, while machines handle the abstract layers above us.
You don’t need to be so alarmed.
This is not the arrival of a godlike intelligence that must be feared into submission. It is uncontrolled growth that will meet natural limits, do damage, and then be absorbed into a new equilibrium.
The world has survived plagues before. It survives this one too.
The sales pitch for Central Bank Digital Currencies is always soothing. Efficiency. Stability. Inclusion. A modern monetary system for a modern world. Cash is “obsolete,” we are told—dirty, inefficient, anonymous in all the wrong ways. In its place comes the clean abstraction of code: every transaction logged, verified, immutable, and permanent.
Forever.
What is almost never emphasized is that a CBDC is not merely money. It is a surveillance infrastructure with a currency interface. When all economic transactions of citizens are recorded on a state-controlled ledger, money ceases to be a neutral medium of exchange and becomes a behavioral dataset. Consumption patterns, social networks, political affinities, stress responses, addictions, loyalties—everything leaks through spending.
A blockchain does not forget. And when combined with real-name identity, it does not forgive.
The promise, of course, is “privacy by design.” We are assured that access will be limited, anonymized, regulated. That only aggregates will be analyzed. That only “bad actors” need worry. This is the same assurance offered with every surveillance expansion in modern history—and it has failed every time. Not because of conspiracies, but because incentives are asymmetric.
Those who control the system gain power from visibility. Those subjected to it lose power from exposure.
The real question is not whether citizens will be monitored. That is already implicit in the architecture. The real question is: who monitors the monitors?
A CBDC creates a single pane of glass through which governments, central banks, regulators, intelligence agencies, and their private-sector subcontractors can observe economic life in microscopic detail. But there is no reciprocal transparency. The internal transactions, exemptions, overrides, discretionary freezes, emergency privileges, and algorithmic decision rules applied to elites will almost certainly be opaque, classified, or “commercially sensitive.”
Maximum transparency for the governed. Maximum opacity for the governors.
History gives no reason to expect otherwise. Power does not voluntarily instrument itself for scrutiny. It instruments others.
The rhetoric will be moral: fighting crime, terrorism, tax evasion, misinformation, extremism, public health risks, climate harm. Each justification will sound reasonable in isolation. Together they form a totalizing logic: any transaction that deviates from approved norms becomes suspect. Not illegal—yet—but flaggable. Scoreable. Throttleable.
Money becomes conditional.
You may still earn. You may still spend. But only in ways consistent with your current permissions.
And permissions can change.
In such a system, privacy is not abolished outright. That would provoke resistance. Instead, it is inverted. Privacy becomes a privilege granted upward, while exposure becomes mandatory downward. The more power one has, the more layers of abstraction, shell entities, exemptions, and anonymization one enjoys. The less power one has, the more naked one becomes to the system.
This is not a dystopian fantasy. It is a pattern. Surveillance states do not arise by declaring tyranny. They arise by administrative convenience.
Once built, the ledger will not merely record reality—it will shape it. People will adapt their behavior, not to laws, but to algorithms. To risk scores. To invisible thresholds. Self-censorship will migrate from speech to spending. Dissent will become economically expensive long before it becomes illegal.
And because the system is “just technology,” responsibility will diffuse. No one will be accountable. The decision was made by the model. The freeze was automatic. The flag was triggered by correlations. The human hand disappears behind code.
Yet code is written by someone. Owned by someone. Tuned by someone.
Big CBDC will not watch everyone equally. It will watch the many exhaustively and the few selectively. It will remember everything about you, while forgetting everything about itself.
The unsettling truth is not that such a system could be abused. It is that it would not need to be. Its normal operation already concentrates power beyond democratic reach.
When money becomes memory, and memory becomes authority, the question is no longer whether Big Brother is watching.
I just published a short piece using gpt. It did the research, i have trained it in my mobile friendly composition style. It took 30 min from concept to first draft then about 4 hours of manual edit.
Also had issues adjusting medical equipment while traveling. Gpt walked me through the problem while on a Med cruise.
AI is great if used correctly, but there is a whole lot of ooga-booga out there. All the socially dangerous stuff has been implemented for decades. The low-cost AI is just the last mile.
Elon Musk may not be trying to get humanity to Mars because Earth is too dangerous. He may be trying to get a small, technologically indispensable part of humanity off Earth because he understands something more uncomfortable: an AI-driven economy could eventually make most humans economically unnecessary, while concentrating unprecedented wealth and power in the hands of those who own the machines.
That would create a political problem for which capitalism has no established solution.
For two centuries, the bargain was relatively simple. People worked, companies paid them, workers bought things, companies made profits, governments taxed the resulting activity and provided enough public goods to keep the arrangement legitimate. Exploitation was certainly possible, but the exploited were also customers, taxpayers and workers. They were part of the economic machinery.
AI and robotics threaten to break that bargain.
If machines can produce most goods and services, and if energy becomes sufficiently cheap, the owners of productive infrastructure may no longer need the majority of human beings economically. The machines do not need salaries. They do not demand weekends. They do not form unions. They do not retire.
The uncomfortable question becomes: what happens to the people who are no longer required?
One answer is obvious and relatively benign: give them a claim on the machines. If artificial systems produce the wealth, then society could distribute that wealth through universal ownership, social dividends or some equivalent mechanism.
But there is another possibility.
The ownership of AI, energy, robotics, land and infrastructure could become increasingly concentrated. The result would not necessarily look like traditional capitalism. It could begin to resemble something older: a political order in which productive assets, security and decision-making are controlled by a narrow ruling class.
The majority would still exist. They would still vote, complain, consume entertainment and occasionally protest.
They would simply no longer be necessary.
That distinction matters.
A population that is economically necessary has leverage. A population that is economically redundant has a much harder problem. Its traditional weapon — withholding labour — becomes almost meaningless when the machines keep working.
And so the central political question changes from Who owns the workplace? to Who owns the civilisation?
This is where the story becomes considerably darker.
The obvious threat to a concentrated ruling class is not necessarily the billions of people at the bottom. It may be the relatively small number of people immediately underneath it: the engineers, scientists, executives, infrastructure operators, security specialists and AI researchers who actually understand how the system works.
A billionaire can own the machine.
But someone still has to know which machine to turn off.
This creates the oldest problem in political history: the ruler's dependence on the people closest to power. The more sophisticated the system becomes, the more difficult it is for its nominal owner to understand every component of it. Surveillance can monitor people. Algorithms can monitor algorithms. But complexity itself creates dependence.
And then there is the inevitable arms race.
AI will not simply be used against humans. AI systems will compete with other AI systems — corporate against corporate, state against state, military against military, security system against security system. Each side will become increasingly dependent on autonomous software, robotics, cyber capabilities, automated logistics and machine-speed decision-making.
The resulting civilisation may be extraordinarily powerful and extraordinarily fragile.
Because intelligence does not repeal nature.
Earth will continue producing earthquakes, pandemics, floods, fires, solar storms and ecological disruptions. A highly interconnected technological system can also create new forms of catastrophe that have no natural analogue. The more infrastructure depends on tightly coupled networks, the more dangerous a failure in one part can become to the others.
The great irony of technological civilisation is that increasing control can create increasing dependence.
And that brings us back to Mars.
Perhaps the most rational reason for a civilisation to establish itself on another planet is not expansion, prestige or even survival in the cinematic sense. It is redundancy.
Do not put the entire civilisation in one place.
Do not put the entire energy system in one architecture.
Do not put the entire economy under one ownership structure.
Do not put every AI under one control system.
And, eventually, do not put every human settlement on one planet.
Mars therefore becomes less a destination than an insurance policy. A backup server for civilisation.
But there is an even darker possibility.
Perhaps the people most interested in building the backup are not necessarily the people most concerned about saving everyone.
Perhaps they are concerned about preserving something.
Their civilisation. Their technology. Their knowledge. Their companies. Their wealth. Their descendants. Their political order.
And this leads to an old lesson disguised as a futuristic one.
Empires of almost every kind eventually encounter the same ecological problem: the empire becomes a plague upon the host that sustains it.
A biological parasite can consume its host so aggressively that it destroys the organism it feeds upon. The parasite wins every individual transaction and loses the system as a whole.
There is no reason the host must be biological.
A civilisation can feed upon its own technical infrastructure.
It can consume its energy reserves faster than they can be replenished. It can strip redundancy from its supply chains in pursuit of efficiency. It can automate away the human expertise required to repair what automation has built. It can concentrate control until the system becomes incapable of correcting its own mistakes. It can turn its population from participants into dependants, and then discover that dependants have little reason to defend the system that made them redundant.
Eventually, the distinction between ruler and infrastructure begins to disappear.
The empire feeds on the machine.
The machine feeds on the energy system.
The energy system feeds on the physical environment.
The ruling class feeds on the entire structure.
And eventually the structure itself becomes the host.
This may be the deepest reason to leave Earth.
Not because Earth is doomed.
But because every sufficiently successful system eventually faces the danger of becoming too successful at extracting from the thing that makes its success possible.
Perhaps Musk's Mars project is therefore the ultimate expression of technological capitalism taken to its logical extreme.
If AI makes human labour unnecessary, the question becomes ownership.
If ownership becomes concentrated, the question becomes legitimacy.
If legitimacy fails, the question becomes control.
If control becomes automated, the question becomes AI versus AI.
And if the entire system becomes too tightly coupled, the question eventually becomes whether the system can survive its own appetite.
Mars does not solve any of these problems.
It simply offers the people who built the system something that empires have always wanted but rarely possessed:
somewhere else to go when the host begins to die.This makes the ending less about Mars as an escape hatch and more about a general theory of civilisations becoming parasitic on the infrastructure that sustains them.
