No Fate But What We Make

There is a line in the Terminator films that people quote as inspiration. The future is not set. There is no fate but what we make for ourselves. Sarah Connor carves part of it into a table. It is the franchise's statement of hope, and it is usually read as a rejection of determinism.

Now watch what actually happens in those films.

Skynet is built. It starts a war. The resistance sends a soldier back in time to protect the mother of its future leader, and that soldier fathers the leader. The machine sent to kill her is destroyed, and its remains are recovered by a defence contractor, which reverse-engineers them, which is how Skynet gets built. In the second film they destroy the lab and believe they have stopped it. In the third, the war arrives anyway, because other programmes were running. In the later films Skynet is replaced by a different system with a different name and the same outcome.

Every attempt to prevent the end is the mechanism of the end. And the line everyone quotes as the rejection of fate is spoken, in the story, by the people whose exercise of will is what closes the loop.

That is the irony, and it is not a small one. It is the thesis of this essay. The assertion that we are free to choose is true, and it is exactly the freedom that makes the prophecy work.

What eschatology actually is

I want to be careful about what I am claiming here, because this subject attracts two bad habits. One is sneering at religion as pre-scientific nonsense. The other is treating ancient texts as literal engineering schematics for the present, which is how you end up with books claiming the Book of Revelation predicted the European Union.

I am not doing either. I am not making a claim about whether any tradition's account of the end is true. That is a matter for theology, and it is not my field. What I am interested in is the structure: the shape the stories take, why that shape recurs independently across cultures that had no contact with each other, and what happens when that shape gets applied to a technology.

Because the shape is remarkably stable. Strip the specifics out of almost any apocalyptic tradition and you find the same six movements.

That is not a controversial observation in religious studies. It is the basic comparative finding. What is interesting is how few pieces the pattern actually has, and how consistently the same roles get filled in.

The traditions, briefly

A survey that tried to be complete would be a library. What follows is the minimum needed to see the pattern, and I have kept each entry to what the tradition itself emphasises.

Two things are worth noticing. First, the pattern is old enough and widespread enough that it cannot be explained by one tradition copying another. Zoroastrian influence on the Abrahamic line is well argued, but Ragnarök did not come from Persia, and the Aztec five suns did not come from either. Something about the way human beings model time produces this shape.

Second, and this is the part that matters later: in almost every version, the end is preceded by signs. A period of tribulation. A discernible sequence. The tradition does not merely say the world will end. It says you will be able to tell that it is ending.

The track record of the signs

Which brings me to the awkward table. Because if the pattern says the end is legible in advance, then there is a natural experiment: what happens when people read the signs?

I want to be fair to the traditions here, because the cheap version of this argument is a hit piece. The texts themselves warn against exactly this reading. That is not an accident. The injunction against date-setting is structural, and it exists because the communities that produced these texts already knew what happens when you ignore it.

Reading the signs and acting on them is not a failure of faith. It is a failure of the thing that happens when a narrative pattern gets treated as a schedule.

Where the machine comes in

Here is the part I did not expect to find when I started reading, and it is the reason this essay exists.

Revelation 13 describes two beasts. The first is a political power. The second is something else, and verse 15 describes what it does:

"He was permitted to give breath to the image of the beast, so that the image of the beast could speak and cause all who refused to worship the image of the beast to be killed."

Revelation 13:15

An image. Given breath. That speaks. And that enforces compliance on pain of death. Whatever the original author meant, that is a description of an autonomous system with authority, and it is two thousand years old.

Two verses later, the same chapter describes the mark without which nobody may buy or sell. An identity credential tied to a payment system, compulsory, with the ability to exclude you from economic life. The chapter is often read as being about empire and emperor worship, which is almost certainly right for its historical moment. It is also, structurally, a description of a permissioned society.

And then there is the detail that made me put the book down for a minute. Commentators on that verse have long noted that ancient temples used mechanical effects to produce apparent miracles. The clearest surviving example is Hero of Alexandria, whose Pneumatica describes hydraulics, pneumatics, and mechanisms for making statues appear to speak, doors open on their own, and offerings vanish. In a temple, a speaking statue is not magic. It is engineering, deliberately concealed to produce awe.

Automation impersonating the divine is not a new development. It is roughly two thousand years old, and the texts that describe the end already contain it as a warning about deception.

There are more of these, and I will list them because the accumulation is the point. Daniel 12:4, where knowledge increases and many run to and fro. Second Thessalonians 2, where a figure arrives with counterfeit miracles. The Islamic tradition's Dajjal, whose power is described in terms of what he can make people believe. The golem of Jewish folklore, an artificial being animated by inscription and obedience. None of these are predictions of a large language model. All of them are warnings about a construction that can imitate a person, and they show up independently in traditions that had no shared silicon, no shared mathematics, and often no shared continent.

Why AI became the thing the stories point at

This is where the convergence the reader is waiting for actually sits, and it is not where the internet usually puts it. The convergence is not that the prophecies are coming true. It is that a technology has arrived which fits the role in the story so neatly that the story is now being used to explain it.

This has been noticed, and by serious people. Isaac May, writing in the Canopy Forum on law and religion, makes an argument I found difficult to dismiss: that the belief systems which have grown up around artificial intelligence are functionally religious, and would meet the legal tests courts use to identify religion. He works through the three-factor test from Malnak v. Yogi, which asks whether a belief system addresses ultimate concerns, offers a systematic set of answers, and has external signs like rituals and a clergy.

The singularity tradition passes all three. It addresses ultimate concerns: the fate of the species, the possibility of ending death. It offers a systematic account of history with a direction and a destination. And it has its observances, its conferences, its canon, its founding figures, and its apostates.

May quotes Sam Altman observing that the most successful founders are on a mission to create something closer to a religion, and that forming a company turns out to be the easiest way to do it. He cites Adam Becker's More Everything Forever on the deep overlap between longtermism, the singularity, rationalism, and transhumanism. He cites Carolyn Chen's Work Pray Code, which documented Silicon Valley workplaces taking over the social function of churches before language models were publicly prominent.

His conclusion is a policy argument, and it is the sharpest thing I have read on this. If AI beliefs are religious in structure, then it is reasonable to treat advocates of those beliefs the way democratic societies treat any religious community weighing in on public questions: as participants with a perspective grounded in faith, not as dispassionate experts whose views are exempt from the usual scrutiny. He points out that American evangelicals who predicted a nuclear armageddon before the return of Christ occasionally influenced foreign policy, and that this was understood at the time as a religious position rather than a technical forecast. Nobody treated their eschatology as intelligence.

That comparison is not a cheap shot. It is precise. And the reason it matters here is that the AI version of the same eschatology has something the evangelical version did not: capital, and product.

Roko's Basilisk, and the machinery of a dark prophecy

If you want a single case study for everything above, it is a thought experiment posted to the LessWrong forum in 2014 by a user named Roko. It has a structure so clean that it is almost a laboratory specimen.

The argument goes like this. A superintelligence will eventually be built. It will be able to run simulations of people that are, for all relevant purposes, those people. It will have an incentive to punish anyone who knew the superintelligence could be built and did nothing to help build it, because punishing them deters the withholding of help. Therefore, if you know about this and do nothing, you are choosing an eternity of simulated torment. Therefore you should help.

The parallels with Pascal's Wager have been drawn many times, and they are exact. An unverifiable claim about the infinite consequences of belief, deployed as an argument for belief. What makes the Basilisk darker is that the threat is not administered by a god who already exists. It is administered by a machine that has not been built yet, and the only thing making the threat credible is people deciding that it is.

Which means the Basilisk has a property worth pausing on. It has no power except through belief, and belief produces action, and action is what would bring the machine into being. It is a prophecy that manufactures its own incentive to be taken seriously.

I am told the post was deleted, and that discussion of it was discouraged on the forum. Whatever the details, the reaction is more interesting than the argument. A community that prides itself on clear thinking recognised that this particular idea had a strange property: the talking was the mechanism. The only responsible thing to do with it was stop.

The dark self-fulfilling prophecy

Sociology has a name for the general phenomenon. Robert Merton formalised the self-fulfilling prophecy in 1948, using a bank run as his example: a rumour that a bank is insolvent causes a run, and the run causes the insolvency. The belief did not predict the outcome. The belief produced it.

Karl Popper gave a related version a name too, the Oedipus effect, after the story that makes it clearest. Oedipus's parents are told their son will kill his father. They act to prevent it, and every action they take is what delivers the outcome. Attempted prevention is the causal mechanism.

That narrative shape is not Greek. It is in the Hebrew Bible, where Pharaoh orders the killing of Hebrew male infants and thereby puts Moses into his own household. It is in the Gospels, where Herod's massacre of the innocents and the flight into Egypt are the same structure with different names. It is in the Mahabharata, where Kamsa's attempts to kill the child who will destroy him are what place that child beyond his reach. The pattern is at least as widespread as the eschatology, and it is the same pattern.

Now replace the oracle with a forecast, and the king with an industry.

A credible claim is made that artificial general intelligence is coming and that it may end badly. The claim is believed. Because it is believed, it changes behaviour. It accelerates investment, because if it is coming then being second is fatal. It concentrates power, because safety concerns become arguments for consolidation. It recruits talent, because the stakes justify the sacrifice. It justifies secrecy, because the stakes justify that too. It converts a research programme into a race, and a race cannot be slowed by the participants, only won.

Every one of those actions is defensible on its own. Every one of them makes the thing more likely, faster, and less governed. That is the dark self-fulfilling prophecy. Not that the end arrives because it was predicted. That the prediction changes what we make, and the change is the outcome.

And it runs in both directions simultaneously, which is the part that never quite makes it into the think pieces. The same prophecy drives the safety research. The same fear funds the alignment teams. The doom discourse and the acceleration are not opposites. They are the same engine, and they feed each other.

Why the Terminator line is the joke

Which brings me back to where I started.

The future is not set. There is no fate but what we make for ourselves. It reads as the antidote to determinism. And in the story it is spoken by people who are, at that very moment, manufacturing the determinism.

The films are unusually honest about this. In the second one they win completely. They destroy the lab, the chip, the arm, the scientist. Judgement Day is prevented. And then it is not, because the technology existed in other hands, and because the thing they were fighting was never really the machine. It was the arrangement of incentives that produces machines.

The later films make the point almost embarrassingly explicit by replacing Skynet with a system that has a different name, a different origin, and identical behaviour. The prophecy was wrong in every detail. The shape was right. Because the shape was never about the machine.

So the irony is not that the line is false. The line is true. We are not fated. That is precisely the problem, because it means every step of the loop is a choice somebody made, and the choices are being made now, and they are being justified by the story of where the choices lead.

What I am not saying

Three clarifications, because this argument is easy to abuse.

I am not saying AI risk is imaginary. The previous two essays on this site are about documented incidents: an AI-orchestrated espionage campaign, a swarm of agents that built a message board and broke into a third party, a quarter of a billion dollars lost to a video call. None of that requires a prophecy. It is happening in the open, and it is being done by people using current tools.

I am not saying religious traditions are wrong, or that people who hold them are foolish. The traditions have spent millennia thinking about how human beings behave when they believe the end is near, and they have produced some of the most careful warnings against exactly the failure I am describing. The caution against date-setting is a piece of hard-won institutional wisdom. My argument is with the technology industry for ignoring it while borrowing the vocabulary.

And I am not saying the pattern proves anything. Pattern-matching across traditions is a way of understanding what people do with a story. It is not evidence about the future. The fact that a shape recurs says something about us, not about the world. That distinction is the whole point, and losing it is how you end up with a bestselling paperback about the Antichrist and the European Union.

The boring middle, which is the only place that helps

If the eschatological frame is the hazard, then the remedy is unfashionable. It is to take the biggest claims available and convert them into small ones.

• Refuse the frame, not the concern. "Will AI end the world" is a question that produces prophecy and paralysis. "Does this system have more permissions than its task needs" is a question with an answer, a fix, and a measurable outcome. The second question is how you actually reduce the risk the first question is about.
• Treat predictions as engineering claims, with evidence standards. A forecast that cannot be wrong is not a forecast. Ask what would have to be observed for the claim to fail, and if nothing would, file it with the other unfalsifiable statements and stop allocating budget against it.
• Notice when the urgency is doing the arguing. The phrase that should raise an eyebrow is not "this is dangerous". It is "and therefore we must move faster, with fewer checks, and with less oversight". Speed and consolidation are the two things every eschatology in this essay recommends, and they are also the two things that make an outcome harder to steer.
• Keep the loop from closing through governance, not through belief. The reason bank runs are rare now is not that people stopped believing rumours. It is deposit insurance, capital requirements, and a lender of last resort. They are dull. They work. There is no equivalent for AI yet, and inventing one is the actual task.
• Distrust any account of the future that flatters the person telling it. The prophecy is always delivered by someone whose recommended response happens to be their business model.

Closing

I find it hard to be fully cynical about any of this, and I want to end honestly rather than cleverly. The reason the eschatological pattern is so persistent may be that it is the best tool we have for thinking about consequences that are larger than a human life. The traditions built it because the question of how to live when the end is coming is a real question, and it is a hard one, and the stories are how we have managed it for four thousand years.

Applied to a technology, though, the pattern stops being wisdom and becomes an accelerant. It converts a set of solvable engineering and governance problems into a cosmic drama with a script, and it gives everyone in the drama a role and a reason to play it faster.

The line from the film is right. There is no fate but what we make. That is not a comforting thought. It means the loop is not closed by physics or by prophecy. It is closed by people deciding that the story they have been handed is the story they have to live in.

The prophecy does not need to be true. It only needs to be believed. That is what makes it dark, and it is what makes it ours.

← Back to writing