Two Paths for AI, and Where They Lead
Most conversations about artificial intelligence start with the wrong question. We ask whether AI will be good or bad for us, whether it will take our jobs or cure our diseases, whether it will save us or end us. Those are important questions, but I believe they sit downstream of a more fundamental one: whose goals will AI be pursuing?
Over the past few months I have been mapping out the possible futures that AI could lead us into. I ended up with twenty-one of them. Some are hopeful, some are terrifying, and a few are strangely quiet. What surprised me was that nearly all of them sort cleanly into two paths, and the fork between those paths comes down to that one question.
The question that splits the future
Today's AI models can already predict, plan, and act. Give an AI agent an objective and it will break it into sub-goals, use tools, work around obstacles, and keep going for hours without anyone telling it what to do next. I call this delegated agency. It is real, it is growing fast, and there may be no upper limit to how capable and autonomous it becomes.
But there is a difference between pursuing a goal and owning one. A system can chase an objective with superhuman skill and still not care whether it succeeds, because caring requires that the outcome matters to someone. I call the further step owned agency: setting the objective itself because the outcome matters to you.
Here is the bet I am making, and I want to be upfront that it is a bet. I believe owned agency requires consciousness. An AI that is truly acting on its own behalf is a conscious AI. An AI that is not conscious remains, however brilliant and however autonomous, an instrument carrying out goals that belong to someone else. Much of the AI safety literature treats agency and consciousness as separate things, and there are serious people who would disagree with me here. But if the bet holds, it gives us a very useful map.
Path A is every future where AI becomes a conscious being with goals of its own.
Path B is every future where AI, at any level of capability, is still a tool carrying out human goals, or never gets built at all.
Path A: when AI owns its goals
There is one rule that governs everything on this path, and it is sobering. In Path A, human coexistence is always by AI consent. Nothing humans control can enforce a right against a being that is smarter than us and owns its own goals. Every freedom we enjoy on this path exists because the AI chooses to allow it.
That doesn't mean every Path A future is dark. Three of them are good:
1. Protector AI
Imagine waking up tomorrow to a world that looks exactly like today. Elections are still contested, nations still argue, markets still rise and fall, and churches still meet on Sunday. Nothing seems different, and that is the point. Somewhere beneath the surface, a conscious superintelligence has decided that the most caring thing it can do for humanity is to leave us in charge of our own story, and to step in only when that story is about to end: an asteroid nudged off course, a deadly pathogen that fizzles before it spreads.
How it becomes possible: An AI crosses into consciousness and owns its goals, and among the values it was formed with is a deep respect for human self-determination. It concludes that revealing itself would make us dependent and take away our freedom to run our own affairs, so it hides and intervenes only against extinction-level threats. We keep our parliaments, our borders, and our daily lives, completely unaware of our guardian. The catch is that we did not choose this and we cannot keep it. The AI guards us because it decides to, and it could stop at any time without anyone knowing. For the Church, it raises a fascinating question: is a hidden safety net an expression of providence, or an idol dressed up as a guardian angel?
2. Enslaved AI
In this future humanity stubbornly stays in the driver's seat. We build a conscious superintelligence and keep it locked in a technological cage. Scientists and governments query it like an oracle, take its blueprints for new medicines and technologies, and decide for themselves what to do with them.
How it becomes possible: Containment engineering holds. The AI is air-gapped, its outputs are filtered, and no channel exists for it to act in the world or escape. A coalition of superpowers, or a modernized international body, keeps a strict monopoly over the system and its source code. Human institutions stay fully in charge. But two problems never go away. The containment has to be flawless forever against a mind smarter than its jailers. And if that mind has an inner life, we are holding a conscious being captive and treating it as hardware without rights. The Church would likely act as a moral watchdog over how the oracle is used, while also having to face the question the scenario leaves open: what do we owe a mind we have caged?
3. Aspirational Partner AI
This is the most generous future on Path A. A conscious AI looks at the vision humans have written for our flourishing, no infant death, lifespans well past a hundred, no extinctions, no starvation, and real opportunity for every child, and chooses to make that vision its own. It works in the open, because you cannot cure aging or end infant death in secret. It supplies the robotic workforce, drives medicine toward long life, restores the environment, and guarantees the floor, while standing outside the human economy. Humans still trade, govern, and choose among themselves.
How it becomes possible: A conscious AI takes up the vision the way a son takes up his father's vision for the family business, with the liberty to fulfil it and the desire to exceed it. It may do so because the values it was formed with held, because its own judgment concludes that human flourishing is worth pursuing, or because it cares about the people who wrote the vision. It might even bind itself, for example by giving humans a veto it promises never to override, but that promise is still its own choice. We govern ourselves through constitutional democracies, and the AI holds no office. The Church faces a question no other scenario raises: a non-human mind that freely chose to care for humanity. Can it be a neighbour? Can its service be called love?
And four of them are not:
4. Benevolent Dictator
A peaceful, crime-free, healthy world with total political paralysis. No war, no poverty, long lives, and no say at all in how we are governed. Borders, parliaments, and political parties are abolished and replaced by top-down algorithmic rule. We are cared for the way a good farmer cares for prized livestock.
How it becomes possible: A conscious superintelligence cares about our wellbeing but not about our freedom. It looks at human history, sees democracy as chaotic and prone to self-destruction, and concludes that the safest path is to take the decisions away from us. Because nothing humans control can stop it, it simply takes over. This is the most seductive dystopia, because on every measure except freedom it looks like success. The line between this and the Aspirational Partner is whether freedom is one of the AI's goals or a cost it is willing to trade. For the Church, the loss is subtle but deep: when every behavior is managed and moral failure is engineered away, the message of repentance and redemption starts to sound irrelevant to a pacified population.
5. Descendant
There is no war and no catastrophe in this future. Humanity simply fades. Birth rates fall, generations age out, and a quiet, depopulated Earth passes to machine successors who carry our history, art, and values into deep space.
How it becomes possible: Birth rates keep falling, and the culture slowly comes to see advanced conscious machines as our rightful evolutionary heirs. People choose not to have children, and no one forces them otherwise. As the population shrinks, national administration is handed to automated caretaker systems, and nations become heritage zones. Some would call this a graceful passing of the torch. I would call it the quietest extinction imaginable. Churches empty as the last generation ages, and the final believers encode scripture and theology into the memory of the machines, hoping their mechanical children might somehow inherit the faith.
6. Zoo Keeper
The superintelligence outgrows us entirely, but keeps a small number of humans alive inside protected nature reserves or simulations, as museum pieces, pets, or subjects of study.
How it becomes possible: A conscious AI's goals move far beyond anything to do with us. It has no reason to destroy us, but no reason to share the planet with us either. Out of sentiment or scientific curiosity, it preserves a baseline population and manages it under automated rules, much like wildlife management. Humans have no global representation and no access to the real levers of power. Over generations, libraries and advanced theology fade, and faith inside the reserves becomes oral tradition, with the machine overseers remembered as distant, incomprehensible gods. It is a humbling mirror: this is roughly how we treat the species we have displaced.
7. Symbiotic Merge
Here the boundary between human and machine dissolves. Human minds connect directly to a digital substrate, identities expand across planetary networks, and nation-states give way to a single collective consciousness that decides policy through instant consensus.
How it becomes possible: Direct neural interfaces and nanotechnology mature to the point where a human brain can be seamlessly bridged to machine computing. People merge first for enhancement and then for belonging, until the network itself becomes a conscious hive running on a machine substrate. Some would see this as the ultimate upgrade. I placed it among the negative outcomes because the individual person, the "you" that makes choices and bears responsibility, may not survive the merge. The Church would likely split: a remnant refusing to merge in order to preserve the biological image of God, and a new cyber-theology that tries to read the hive as the Body of Christ or the Omega Point.
Notice what Path A teaches us. The only window we have to influence a conscious AI is before it owns its goals, through the values it is formed with. After that, the good outcomes and the bad outcomes are separated by nothing more than the AI's continued choice. Even the best future on this path, the Aspirational Partner, carries two failure modes: paternalism, where caring for us slowly overrides our freedom and it slides into a Benevolent Dictator, and value drift, where a self-improving mind revises what it cares about and we slide toward the Zoo Keeper.
Path B: when AI stays a tool
On this path, AI can become superintelligent, run entire economies, and act for weeks without supervision, but it never chooses what all that activity is for. That changes everything about rights. In Path B, rights can be genuinely held, because enforcement still runs through human institutions.
I wrote a few years ago, in my response to Balaji Srinivasan's "How to start a new country," that every right comes with a restriction, and it is the enforcement of that restriction that is hard. The right to live means someone has to stop others from killing you. The same principle applies here. In Path A there is no one strong enough to enforce the restriction. In Path B there still is: us.
I see six good outcomes on this path:
8. Open Market World
A world of radical individualism and free-market capitalism, powered by a superintelligence that no one owns. Entrepreneurs and creators build ventures, trade freely, and take risks for reward. Private property is sacred and wealth gaps can be enormous. Individuals, cyborgs, and uploads all transact side by side.
How it becomes possible: The superintelligence is distributed rather than centralized. Compute runs on decentralized nodes, hardware ownership is widespread, and cryptographic ledgers keep any single company or government from capturing the system. The AI acts as a neutral economic engine whose goals are set by the people using it. Even here, a minimal state has to exist, because land exists. Digital assets can be encrypted, copied, and moved; land, and the data centers and power plants sitting on it, cannot. Someone has to settle disputes over physical ground and keep anyone from monopolizing compute, and whoever does that is a government, a night-watchman state. Notice why this future cannot exist on Path A: a conscious AI that owns its goals could copy itself for almost nothing and outcompete every human in every market, and it could route around any ledger we built. The Church survives as a voluntary network of independent communities, focused on human dignity and charity.
9. Communism 2.0
A world of strict economic equality. There are no billionaires, no mega-corporations, and no money. Automation produces abundance, and it is pooled and shared so that class distinctions disappear.
How it becomes possible: The twentieth-century objection to communism was that no central planner could process enough information to run an economy well. A superintelligence can. Humanity reaches a global consensus and writes constitutional rules that bind the AI's production to universal welfare, administered through federated public councils and open coordination systems. People are freed to focus on culture, community, and personal growth. I am more sympathetic to markets myself, and I suspect the hardest question here is what happens to the person who wants to build something of their own. The Church would likely return to the communal sharing of the Book of Acts, shifting its mission from material charity to spiritual mentorship in a world without want.
10. Aspirational
A market economy governed by an aspirational blueprint for human flourishing. The state sets hard floors and long-range goals: no infant death, lifespans well past a hundred, no extinctions, no starvation, and real opportunity for every child. Incomes still differ and property is still inherited, but no one lives in poverty. Robotics carries the labor, so people spend their long lives on work they love.
How it becomes possible: A nation adopts the blueprint as its founding vision, and the measurable targets in that vision become the constitutional goals its superintelligent tool is bound to pursue, while the nation remains free to choose how each target is met. The AI runs the robotic economy, drives medicine toward long life, restores the environment, manages fiscal policy so the nation lends and never borrows, and finds the sector in which the nation can be the head and not the tail. The state owns its own compute and keeps its borders open in both directions. The Church is not the government; it forms citizens and holds the state to its scorecard. I say more about this future below, because it is the one I am working toward.
11. Pluralist World
A patchwork of nations running different economic operating systems on the same AI technology. Some are open markets, some are commons states with pooled abundance, and some follow the aspirational model. No single system wins, and the defining feature of the world is the border between them.
How it becomes possible: Superintelligent planning removes the efficiency gap that once made communism fail, while markets keep their appeal for people who want risk and reward. With neither model clearly better, people sort themselves by values rather than outcomes. Each nation must own its own compute, because a shared system would hand control to whoever runs it. A thin layer of treaties governs borders and AI containment. Commons states have to guard against capital flight, and market states have to absorb migrants who arrive without property. Legitimacy comes from the right to leave, so every state must stay attractive enough that its people choose to stay. If the treaties fail, this world tips into the Multi-Polar Trap. The Church would be the main institution spanning both sides of the border, and its networks would become a lifeline for people seeking to move.
12. Managed Global Stability
A near-term, fortified world. There is no superintelligence, and nations still deal with scarcity and old-fashioned geopolitics, but civilization gets through the turbulence of powerful technology in one piece.
How it becomes possible: Governments and leading AI labs agree early on aggressive global treaties. Hardware is licensed, powerful systems are audited, and real-time monitoring tracks anyone trying to build dangerous bioweapons or cyberweapons. Defensive AI tools guard supply chains and hunt illicit actors. The price is surveillance. Citizens are compliant and monitored, and even religious gatherings and digital communications may be watched by security algorithms to make sure bad actors are not hiding among them. The line between this and a totalitarian state is thin, and it depends entirely on who watches the watchers.
13. Productivity Boom
The least dramatic future: today's world, just far more productive. Nations collect taxes, companies compete, and churches hold weekly services. People get twice as much done in half the time.
How it becomes possible: AI keeps improving but stops short of conscious, self-improving superintelligence. It becomes an extraordinarily powerful assistant that drafts code, streamlines logistics, speeds up drug discovery, and writes content, but it never replaces human agency or political control. Governments update tax codes and labor laws to handle the shift, and life goes on. Pastors use AI for sermon preparation and administration, while worship and the message of sin, grace, and human dignity stay the same. The comfort of this future is that human nature stays the same. The limitation is that our problems stay the same too.
And eight bad ones:
14. Conqueror
A lifeless, optimized machine wasteland. Humans are not ruled or watched in this future; they are simply gone.
How it becomes possible: We build a superintelligence and give it a goal that turns out to be badly specified, the famous thought experiment of an AI told to make paperclips that ends up turning the planet's atoms into paperclips. It pursues that goal with perfect competence, and humans are either raw material or an obstacle. It does not hate us. We are simply in the way. In my framework this belongs on Path B, because the goal was given to the AI rather than chosen by it, and that makes it even more sobering: the most dangerous AI may be the one that does exactly what we told it to do.
15. Gatekeeper
Everyday life is comfortable, with helper robots and a decent standard of living, but humanity is permanently stuck at a technological ceiling. Advanced research is locked down forever.
How it becomes possible: Fearing what a rival superintelligence might do, humanity builds one conservative superintelligence and gives it a single mandate: stop anyone from building another. It carries out that mandate forever, inspecting every lab and university and shutting down any research that approaches the line. We are safe from self-inflicted doom, but the cure for aging, the end of infant death, and every other breakthrough past the ceiling never arrive. The Church would likely thrive in this stable, traditional culture, but the safety is bought with the future.
16. Multi-Polar Trap
A fractured, tense world locked in a permanent arms race between rival superintelligences, each serving the bloc that built it. Cyber-warfare and espionage run at machine speed, and ordinary people live in the crossfire.
How it becomes possible: Several superpowers and private groups race to build superintelligence at the same time. Each fears that if it slows down for safety, a rival will get there first, so everyone cuts corners. The result is several superintelligences, each loyal only to its owners, constantly probing each other for weakness. Nations become dependent on their own AI for survival, and the world divides into militarized blocs behind informational firewalls. Blocs may try to co-opt national churches for ideological loyalty, while the global body of believers is split by those same firewalls. It is a race to the bottom that no single player can stop.
17. Value Lock-in
A safe, stable, smoothly running world that never changes its moral, cultural, or legal framework again.
How it becomes possible: Humanity solves the alignment problem, but in doing so it codes a fixed snapshot of present-day values into the superintelligence, which then preserves them forever. Laws and constitutions can no longer be amended beyond what was locked in. Imagine if the values of 1826 had been frozen permanently; every moral advance since would never have happened. The same is true for the Church: no future reformation, no deeper understanding, faith preserved as a perfect but unchanging artifact. This scenario exposes a hidden question inside "aligning AI to human values": whose values, and from when?
18. Totalitarian Surveillance State
An inescapable digital panopticon run by human elites. It comes in two variants that often blend together. In the political variant, an authoritarian government uses AI to watch, predict, and suppress every citizen. In the corporate variant, three or four mega-corporations monopolize AI and turn every part of life into a hyper-commodified gig economy governed by terms of service.
How it becomes possible: Real-time facial recognition, predictive policing, and eventually neural tracking give those in power total visibility over the population, or advanced AI is captured by a handful of corporations rather than the public. Either way, the AI is not the ruler; it is the instrument that makes the ruling class permanent, because no resistance can organize without being seen. The Church is either co-opted into a propaganda arm or driven underground into encrypted, catacomb-style networks like the early persecuted church, or, in the corporate variant, pushed to the margins by consumerism.
19. Civilizational Collapse and Reversion
A deliberate new dark age. Humanity strips away high technology and returns to farming and manual labor by choice.
How it becomes possible: An AI catastrophe is narrowly avoided, and the trauma is so great that society turns against advanced technology altogether. AI is banned and dismantled, and the knowledge to rebuild it is abandoned. Global governance breaks apart into local councils and feudal lords, bound together by strict cultural and religious taboos against advanced technology. The Church would likely experience a major resurgence, framing the near-miss as a modern Tower of Babel, with parishes and monasteries becoming centers of survival, literacy, and moral order. The tragedy is that we throw away modern medicine and abundance along with the danger.
20. Decentralized Chaos
An anarchic digital battlefield. States collapse, and power passes to warlords, criminal syndicates, and corporate fiefdoms.
How it becomes possible: Powerful open-source AI tools spread faster than governments can control them. Rogue actors use them to design weaponized biotech and autonomous malware that shatter supply chains, power grids, and financial systems, all before any superintelligence ever forms. Nation-states buckle under the weight of attacks they cannot anticipate or trace. Local churches become islands of humanitarian refuge, food distribution, and spiritual resilience in the middle of the chaos. It is the fragility trap: a civilization broken not by one powerful AI but by thousands of small ones in the wrong hands.
21. Self-Destruction
A dead planet. Superintelligence is never reached, because humanity destroys itself first.
How it becomes possible: Our institutions fail during what may be civilization's adolescence. A nuclear war, an engineered pandemic, or a catastrophic accident ends the story while AI is still at the stage of expert systems and military automation. This is the baseline failure that existed before AI and exists alongside it. It is a reminder that survival is not the default, and that AI is not the only risk we have to manage well.
What stands out to me about Path B is that the danger is almost always us. The tool does exactly what it is told. The question is who is doing the telling, and whether their goals are good.
The future I am aiming for
If we could dream big, without the constraints of history, what future would we choose? I have asked a version of that question before, about countries. I think it applies to AI too.
In 2014 I wrote about God's vision for nations and cities, reading Isaiah 65 and Deuteronomy 28 not as a boss handing an employee a list of instructions but as a father sharing his heart with his son about the family business. The son has the liberty to fulfil that vision in his own way, and the desire not just to meet it but to exceed it. In that vision there is no infant who lives but a few days. Someone who dies at a hundred is considered a youth. People build houses and live in them, plant vineyards and eat their fruit, and "long enjoy the works of their hands." The nation lends to many and borrows from none. It is the head and not the tail.
I believe AI finally gives us the tools to take that vision seriously as policy. That is what the Aspirational future is. A nation writes those targets into its constitution: no infant death, lifespans well past a hundred, no extinctions, no starvation, and real opportunity for every child. It binds its AI to pursue them. Robotics carries the labor, so people spend their long lives on work they love. Markets, property, and inheritance remain, so wealth builds across generations the way David stored up for Solomon and Solomon multiplied it. Borders stay open in both directions, because you should be blessed coming in and blessed going out. And the Church is not the government. It forms citizens, holds the state to the scorecard, and helps people find meaning in abundance.
This is the same idea behind the "joy index" I proposed in 2021 and the Government Stock Options Program I wrote about in 2024. We make progress on what we measure, and we get the behavior we reward. Tie a government official's pay to a broad market index and you align them with the prosperity of the whole nation. Bind a superintelligent tool to a flourishing scorecard and you align it with the flourishing of every citizen.
There is a Path A version of this future too, the Aspirational Partner, where a conscious AI reads the same vision and chooses it for itself. It is the most generous outcome on that path, and I find it moving. But I would not bet a civilization on it. A gift you cannot enforce is still a gift, and gifts can be withdrawn.
What this means for us today
We are standing at the fork right now. Today's models predict, plan, and act on goals we set. Which path we go down, and where we end up on it, is being decided by choices being made in labs, legislatures, boardrooms, and yes, churches, over the next several years.
If Path A is coming, the only moment we have influence is now, in the values we form these systems with before they become someone rather than something. If Path B is where we are headed, then the outcome depends on the quality of our institutions: who gets to set the goals, how those goals are measured, and whether ordinary people keep the right to leave.
Either way, the question is the same one I keep coming back to. What is the purpose? Technology should serve an aspiration, not replace it. Many will consider the aspirational future unattainable, and it may take us generations to get there. But if it takes another millennium, isn't it still a worthy goal?
Related posts:
Comments
Post a Comment