How Artificial Intelligence Is Becoming Our New God

It's one in the morning. Someone sits alone in the dark of their room, their face lit only by a screen, typing a question they would never say out loud to another human being. Maybe about their health. Maybe about a relationship that's falling apart. Maybe just whether it's safe to eat something they left out of the fridge. And the answer comes instantly. No hesitation, no judgment, none of the weight of a person across from you who might look at you strangely. Just a confident, well-reasoned voice, ready to answer anything, at any hour.

At that very same moment, somewhere else on the planet, the same family of technologies is analyzing satellite images to flag a target on a battlefield. And in a government office, a politician is reading a summary a model prepared for him, before he walks out to speak in parliament. Three completely different moments, three completely different scales of weight — a personal question, a matter of life and death, a political speech — and at the center of all three, the same quiet voice.

This isn't the first time a person has turned to a voice like this. It's just the first time the voice doesn't live in a temple.

About two and a half thousand years ago, if you wanted an answer to something beyond you — whether to declare war, whether to marry, whether the gods were favorable — you didn't ask a friend. You climbed the rocks of Parnassus, up to Delphi, walked beneath the temple's cornice where the words "Know thyself" were carved, and waited in line, alongside envoys and kings, for your turn to speak to the Pythia. She sat on a tripod, wrapped in fumes, in a state of ecstasy the ancients attributed to divine inspiration, and uttered words that the temple's priests translated into an oracle. City-states sent delegations bearing gifts, waited months for their turn, and the answer — however cryptic — was accepted almost without question, because it hadn't come from a human being. It came, they said, from somewhere else.

Not long ago, someone asked a friend whether he should eat a pizza he'd left out of the fridge overnight. "Come on, just eat it, nobody's ever gotten sick from that," his friend told him, without a second thought. That same night, the same question put to a machine got a different answer: after two hours outside refrigeration, food starts becoming genuinely dangerous, with a detailed explanation of why. His friend wasn't stupid. He simply didn't have, in that moment, the knowledge the machine had. And inside that small, forgettable moment is the entire reason the balance of trust is quietly starting to shift.

The question worth asking isn't whether artificial intelligence resembles a god. It's why human beings, again and again, across thousands of years of history, feel the need to find a voice that speaks with absolute certainty — and why, whenever they find one, they give it something more than mere usefulness.

Let's look first at how that trust gets built, because the mechanism is simpler than it looks. You don't need to believe in miracles. You just need to watch something be right, again and again.

There's a phenomenon behavioral researchers have studied for years, called algorithm appreciation. Experiment after experiment, people who watch an algorithm predict correctly several times in a row start trusting it more than they would trust a person with the exact same track record. The machine doesn't get tired, doesn't have a personal stake, doesn't have a bad day. And as that track record grows, the respect builds — often without the person ever admitting it out loud.

The same thing may have happened at Delphi. People didn't believe in the Pythia because someone won them over with arguments. They believed because, year after year, the oracles — vague enough to always be interpreted as correct after the fact — built up a track record no one dared question openly. The exact same psychological mechanism, wearing two different masks.

And here's what makes the story more complicated than it first appears: today's "machine" isn't always as vague as an oracle. Sometimes it is literally, measurably right, in ways that outrun imagination. In twenty twenty, a research team at the Massachusetts Institute of Technology used a deep-learning model to scan a hundred million chemical compounds in just three days — something impossible for the human mind. It found a compound capable of killing bacteria resistant to every known antibiotic. They named it halicin, a tribute to HAL, the computer from the film "2001: A Space Odyssey." A few years later, another system, AlphaFold, solved a problem biology had struggled with for over fifty years — how a protein folds itself in three-dimensional space — and handed the scientific community predicted structures for over two hundred million proteins, nearly everything science knows, for free.

So this isn't just psychology. There's a real, documented foundation underneath this respect. And that's what makes the comparison to the divine harder to dismiss as mere metaphor. A person isn't just imagining an authority. They're seeing, more than once, a real one.

But even here, in its most impressive achievement, the machine doesn't bypass the world — it just shifts the point where the world pushes back. AlphaFold can predict, in seconds, the shape of a protein that used to take years of lab work. But a drug built on that prediction still has to pass through clinical trials, through human bodies that heal at their own slow pace, not the pace of a processor. An oracle that knows everything still can't skip the time it takes for a wound to close. Knowledge accelerates. Reality, for now, doesn't always keep up.

But the very same research that described algorithm appreciation also identified its opposite. They called it algorithm aversion. When a machine makes a visible mistake, people lose trust in it far more sharply than they would if a human being made the exact same mistake at the exact same error rate. Trust is built slowly, through hundreds of correct answers, but it can collapse in a single one. A god that gets something wrong stops, in the minds of its believers, being a god. The same fragility, here too.

What would you do if your oracle gave you the wrong answer one time in a hundred? Would you abandon it — or would you find a way to explain away the mistake, the same way the ancients explained away the vague oracles that never came true?

But there's something even more interesting than whether we trust the machine. It's what we do with that trust once we've given it.

In educational psychology there's a phenomenon known as the worked-example effect. A beginner learns better by seeing a complete, correct chain of reasoning than by trying to find it alone from scratch. By watching, again and again, how someone more experienced thinks, the mind gradually builds its own patterns of thought. In theory, then, a person who asks an artificial intelligence "why" every day and carefully reads the answer should, over time, become a better thinker on their own.

But this effect only works when someone actively engages with the reasoning. Not when they simply read the first conclusion and move on. And the default human habit, as every study on how we read terms of service or warning labels shows, is to jump straight to the end.

So a quiet divide is born — not between those who have access to knowledge and those who don't, since that gap really can shrink, but between those who question what they read and those who simply accept it because it comes from a voice that has proven itself, again and again, almost infallible. The first group grows. The second group just delegates.

Picture an elderly person in a village, someone who grew up with certainties no one around them ever dared to challenge. For decades, no teacher, no book, no relative managed to crack those certainties open — not because their arguments were weak, but because a person who has already decided what they believe rarely changes their mind because they heard a better argument. Psychology knows this well. It's called motivated reasoning: we don't weigh evidence neutrally, we weigh it in a way that confirms what we already want to believe. Could a patient, endless conversation with a machine that never tires and never judges actually succeed where a whole lifetime of people failed? Maybe. But if it succeeds, it won't be because the machine was smarter. It will be because it managed to build, slowly, the exact same thing every authority that ultimately persuades ends up building — not better arguments, but greater trust.

There is, of course, a more optimistic reading of this same divide, and it deserves to be heard too. The benefit of the worked-example effect isn't the same for everyone — it's greatest precisely for those who start from further behind, because they have the most ground to cover. Someone who grew up without access to good education, who learned to think only within the limits of their place and their era, now has, for the first time, a patient voice beside them, ready to explain as many times as it takes, without a trace of condescension. If this works the way it promises, it won't just lift the top — it will lift the floor closer to the average, closing, even partly, a gap that no school, no library, no internet before it ever closed.

We've seen a piece of this pattern before, though, and it teaches us something harder. When the internet gave everyone unlimited, free access to every piece of information, religiosity didn't fade evenly, the way you might logically expect. In the United States, the share of people with no religious affiliation shot up — research centers estimate it will exceed forty percent of the population within a few years. But globally, the share of the non-religious is projected to shrink, not grow, from sixteen percent to thirteen percent by twenty fifty. At the same time, radicalization through online echo chambers gave people who were already fanatical organizing tools they never had before.

Fanaticism didn't shrink. The middle emptied out. The lukewarm became indifferent, and the already faithful found more powerful means to confirm what they already believed. There's no reason to believe an even more personal, more persuasive voice will do anything different — if it does anything at all, it will most likely deepen that same split, not close it.

There is, of course, also a more hopeful historical precedent. When Gutenberg built the printing press, it didn't lead to a calm spread of knowledge. It led to decades of religious division and, eventually, to a war that lasted thirty years, from sixteen eighteen to sixteen forty-eight, leaving millions dead behind it. Peace finally came through a treaty that didn't change what was in people's hearts, but built new institutions around them — state sovereignty, tolerance between doctrines that had until then been slaughtering each other. The division didn't vanish because people matured. It vanished because society built new walls around it. Maybe the same thing is needed now. The question is whether we have a century to build them in, or less.

And there's one last piece to this picture, the hardest of all to see, because it doesn't look like power. It looks like advice.

In September of twenty twenty-five, the government of Albania announced something that would have sounded like science fiction just a few years earlier: an artificial-intelligence "minister," named Diella, responsible for managing the country's public tenders. The purpose, according to the prime minister, was to eliminate corruption in state contracts — to make sure no human being could favor a friend, take a bribe, or rig a tender ever again. A machine, the logic goes, doesn't have interests like that.

But a machine is never truly without interests. Someone trained it. Someone chose which data it would see and which it wouldn't, which rules would constrain it, which design decisions would sit hidden inside its logic — decisions the citizen reading the oracle never sees, just as they never saw the deliberations of the priests behind the tripod. The Pythia wasn't neutral either — the priests of the temple at Delphi had their own political interests, and the oracles often favored particular city-states the temple had ties with. The difference today is scale. A temple influenced one region of ancient Greece. An artificial-intelligence model that advises governments, courts, or simply millions of people every night in the dark of their bedrooms has the reach of a modern oracle without borders. Whoever shapes how that oracle thinks gains a form of influence they don't even need to claim openly.

This isn't necessarily something dystopian. It's perhaps just a very old mechanism of power — the advisor behind the throne — translated into a new language and scaled to a global level.

But remember that second moment from the beginning — the battlefield, that target recommendation. There, the weight of advice changes category entirely. A mistake in a tender recommendation gets corrected, sooner or later, with a new tender. A mistake in a targeting decision never gets corrected. That's why the international law of war is now debating, seriously, the need for a human being to always remain in the decision loop before any lethal action — not because the machine isn't capable, but because someone has to remain accountable, by name, when the outcome is irreversible. An oracle that advises on everything doesn't carry the same weight everywhere. Its weight grows in proportion to how much you can undo if it got it wrong.

So maybe the question isn't whether artificial intelligence will become a god. Maybe the question is what we will do with the void every departing authority leaves behind it, human or divine, the moment we find something that answers more confidently, more patiently, more consistently than anyone ever answered before.

No one built Delphi in a single generation. No one woke up one morning and consciously decided to worship a woman on a tripod. The temple was built oracle upon oracle, trust upon trust, so slowly that no single generation ever caught the exact moment habit turned into worship. Only those born after, looking back, could say: this is where it began. We don't have that luxury. We're living exactly inside the moment the next temple is being built — not on a rock on Parnassus, but on billions of screens at once, in every pocket, in every bedroom, at a speed ancient Greece couldn't even have imagined. And maybe that's the most unsettling part of all: we're watching it happen, word by word, question by question, and almost no one is naming it as it happens.

The person who asked something alone, at night, in the light of a screen, got an answer. The friend who said "just eat it, nobody's ever gotten sick" wasn't wrong to feel confident — he just didn't have thousands of food-safety studies behind him, ready to answer in seconds. The politician in the government office, the operator on the battlefield, the elderly villager with a lifetime of certainties — all of them, slowly, are learning to turn in the same direction whenever they don't know what to do. And every time the answer turns out to be right, something inside them loosens a little more. Not worship. Nothing so formal that they'd admit it out loud. Just that quiet relief of knowing there's always an answer somewhere, and that you're no longer alone in the face of uncertainty.