What does the Bible say about AI or artificial intelligence?

What is “artificial intelligence?”

It’s important to understand that when engineers and programmers talk about artificial intelligence, they aren’t actually talking about something that functions like the human brain at all. What we call “artificial intelligence” is not a thinking mechanism, but something more like very complicated guesswork. The better name for programs like ChatGPT is “large language model,” or LLM for short. The essence of an LLM is that it takes a question you ask and tries to predict what combination of words and characters is the “right” response. It's unable to actually understand language—not in the way that humans do. Instead, it learns to recognize patterns across millions and millions of associations and combinations of words and then predicts which string of words is the best response to the combination of words that you sent it. If you interact with such programs enough, it quickly becomes evident that the program isn’t thinking or communicating; it just keeps guessing at responses until it gets the one that satisfies you. Its responses can be outright wrong or even nonsensical, because it isn’t actually thinking.

If you ask ChatGPT a question, for example, it might give you a totally wrong answer, or respond to the question in the wrong way. If you point this problem out, it will try to correct its answer, but again, it’s not because it understands what you’re saying. When a person misspeaks in a normal conversation and you correct them, the speaker recognizes that they misspoke and why. But an LLM doesn’t think. All it’s doing is getting negative feedback on some part of the combination of words it provided, so it jumps to its next best guess as to the combination of words that you want. If you get into a “fight” with it, it becomes apparent that it isn’t an intelligence at all. I kept telling it that its answer was wrong (because it was) and the program kept giving new answers that were all on the same general topic but didn’t really make any sense as answers to my question. The machine was simply jumping around to all of the related combinations of words in its memory until I gave it the feedback that it had succeeded in producing what I wanted to hear.

What is an LLM actually like, then? Weird as it may sound, this sort of process is most like a type of divination—what you might call fortune-telling. Divination is the art of trying to read the future based on certain types of signs. The reason that ancient people believed in this sort of art is because they saw the physical and spiritual worlds as intrinsically connected, and they thought that everything in the physical world was also connected in various ways. If there was a weird event in the sky—like an eclipse, a blood moon, a supernova, a meteor shower, et cetera—they thought that it was a sign of something happening or about to happen on Earth. Even in the Bible, the Star of Bethlehem was a powerful astrological sign that Jesus had been born and God had become man. Reading tea leaves, laying out Tarot cards, palm reading, star-reading, and so on, are all just ways of interpreting a specific dataset according to some model in order to predict the future or understand the present.

A Large Language Model works similarly. It takes a huge dataset of information and creates a model for how all of that data is related. Then it uses that model to make predictions about how words and phrases are related. Rather than call this model an artificial intelligence, you might instead call it a kind of simulated intelligence: it isn’t a mind or an intellect; it just simulates or creates the illusion of one.

While predicting what you want to hear is a little bit like fortune-telling, when you start asking such a machine questions about the weather, the stock market, or world events, then you are certainly engaging in a kind of divination. By creating LLMs, we’ve effectively built the most powerful divination method in human history; and though we might think we’re superior to our ancestors because we have machines while they were reading tea leaves or doing astrology, in reality, we’re all trying to do the same thing: predict the future so we can control it.

Divination and magic in the Bible

While simulated intelligence programs like ChatGPT are not mentioned in the Bible, divination is. The laws of the Old Testament forbid the ancient Israelites from using divination, from practicing magic, and from constructing idols. It wasn't that the ancient Israelites and Christians disbelieved in the reality of magic, demons, or the pagan gods. They absolutely believed in those things, but they were concerned about their danger and wickedness. What danger and wickedness? It has to do with the source of magic and the nature of these spiritual beings.

As I covered in my articles on the book of Enoch and the pagan God Tartak, the ancient Israelites and early Christians believed in the existence of the pagan deities. Their conception of the universe was that the highest God—the Lord of Hosts, the God of Gods, the God of Heaven—created all the lesser gods: the heavenly hosts, the “sons of God,” et cetera. These beings helped God oversee and manage the world, met regularly in assembly to discuss the affairs of the cosmos (see Job chapter 1), and even watched over specific nations. After the fall of the Tower of Babel, the Old Testament records that God divided the nations “according to the number of the sons of God” (Deut 32:8), which was traditionally understood to be a reference to the angelic beings that came to be called the “gods of the nations.” The issue is that, as recorded in the Book of Enoch, these creatures had their own “fall”: they became corrupt slaves of their appetites and created the various religions of the pagan world as a means to obtain worship and sacrifices for themselves.

Divination and magical practices were, in large part, about interacting with these creatures and trying to obtain knowledge from them. Magical healings, wise advice, and especially predictions of the future were all sought after when interacting with these imposter deities. But of course, if such creatures were fallen angels, rebel gods, and usurpers of the highest God, then interacting with them is at minimum going to be dangerous. Such fallen creatures have no interest in being honest, nor do they care about your wellbeing; they want what they want, and so whatever they tell you might be misleading or intentionally dishonest. When you consider the number of absolutely brutal pagan gods who demanded things like human sacrifice and who frequently took out their anger on their worshipers, you can see why the Old Testament law of Moses would forbid any kind of interaction with them.

In the same way, building a massive, mechanized divination machine could have similar dangers. Because the algorithms are designed to tell you what you want to hear, there’s a strange sense in which building an LLM is about making a machine whose whole focus is to fulfill our desires and show us what we want to see. You might think that that sounds great, or that that’s the purpose of technology, but think about it more soberly for a minute: do you always want what’s best for you? I know that I don’t. What about wanting to know the truth? We very often don’t want to know the truth, or the full truth, because it can be painful and may force us to change. It usually takes someone you love to really help you understand where you’ve gone wrong; by default we tend to avoid painful truths and seek only what is pleasant.

One of the big differences between this new kind of mechanical divination and the divination of our ancestors is the question of truth. Divination methods like reading palms, tea leaves, cloud patterns, or the motion of the planets all have something in common: they’re all about the natural world. Because ancient people believed in an enchanted world—the idea that all physical things had spiritual content—they believed that nature itself contained real spiritual truth that you could learn if you knew how to read it properly. To do divination well was about humbling yourself before nature and, in some sense, being a student of nature. In other words, the idea was that truth exists outside of you: it’s out there, in nature, and it’s something you can gain access to if you know how. You might think that these methods don’t work, that there’s no real truth in tea leaves, cloud formations, or the motions of planets. But it’s important to understand why people might think so: namely, ancient people fundamentally believed in truth that was bigger than them and could be known by them.

But with a modern intelligence simulator, the question of truth becomes very muddled. Remember that such a program works by guessing at the right response. It’s not focused on the truth of the things it’s saying, because it doesn’t have a way to verify the truth. How could it? All it can do is guess based on pattern-recognition about what words and ideas are related to each other.

As such, an LLM actually has a strong bias. First is the obvious bias that comes from the people who programmed it. But the second is the way it gets positive feedback for answers that the user likes and negative feedback for answers that the user doesn’t like. This criterion is very different from whether something is true or not.

So what’s the issue? The issue is that we’re building something whose fundamental goal is to tell us what we want to hear. In other words, our desires have been substituted for truth. A simulated intelligence is not an objective source of truth or even of factual information. At a fundamental level, it’s a mirror of our own desires; one that could become so good at predicting what we want that it knows what we want before we realize it. It’s like taking our all of our most basic impulses and urges and then giving a machine absolute power to gratify them, especially if you put an LLM in control of a virtual reality interface with the task of entertaining you in various ways.

Again, such a system is not as good as it sounds. In order to raise ourselves up higher than our biological urges, we have to do all sorts of things like delay gratification, make sacrifices, abstain from pleasure, and think long-term. The state of having all your impulses constantly and instantly gratified is a kind of hell: it reduces the human being to a mere animal and enslaves you in the dungeon of your urges.

The science of idol-making

The concerns we’ve discussed so far assume a more benign vision of this new technology. In fact, the above scenario is the minimum of what this sort of technology is capable of doing to us, especially when combined with virtual reality and the ability to generate any combination of images, sounds, and sensations. But there could be a much darker side to it. I want to end with a final concern, something that comes straight out of the Bible and the experience of the early Church. You might find it a little weird, but I think it’s worth thinking about.

There are many places where the Biblical authors condemn interaction with what they call “idols.” In many modern Christian circles, this language has been used to mean things that you “worship” instead of worshiping God. If you get obsessed with a sports franchise, a hobby, or whatnot, someone might say that it has become your “idol,” that you need to make sure you’re not giving it so much “worship” that it’s displacing God in your life. While it’s important to moderate your behavior and be aware of when things are disrupting your spiritual life, this modern usage was not what the Biblical authors meant when they talked about idols. In the ancient world, making and worshiping idols was a near-universal pagan practice that was understood in a specific, literal way.

Pagan thinkers across most cultures believed in something that we might call the “science of idol-making.” Their understanding was that if you build a physical object such as a statue in just the right way, and you performed certain magical rituals on it, you could actually lure and summon a spirit into the statue and bind it there. It is the same idea as a “genie in a bottle”—a powerful spirit trapped in a physical object.

Hermes Trismegistus, a famous pagan author, explained this “technology” to his son in the ancient document Corpus Hermeticum. He tells his son that their ancestors “invented an art by which they could create gods,” such that they “were able to create idols having both good and evil powers.” Some of these idols, he claimed, had the ability to “foretell the future through lot and divination, or [] provide certain things, and by these means [] come to help human beings, each in his own way.”

When pagans built idols, they created statues of gods or spirits and lured or bound those spirits into them so that they could gain access to the spirits’ powers. You might think of this practice as pure superstition, but if you are a Christian, you should take it seriously. The Biblical authors and the early Christians all believed in and took this practice very seriously—that’s why they condemned it.

What am I saying, therefore, about the advent of AI and LLMs? I don’t want to say anything too strongly yet. I’m not saying that computers are “the devil” or anything like that. It’s just that when I hear that we are, in essence, working very hard to build physical objects in such a way that we can put a kind of “mind” inside of them that can predict the future or tell us what’s best, it sounds a lot like idol-making.