Faith and AI  ·  By John Moelker, a pastor who builds AI

What a large language model actually is, explained for Christians

A lot of the fear and the hype about AI both grow in the same soil, which is not understanding what the thing actually is. So let me do the most useful thing I can as someone who builds these systems: explain it plainly, with no jargon and no agenda, the way I would explain it to a friend across the table. You do not need to be technical. If you can follow autocomplete on your phone, you can understand this.

It predicts the next word. That is the whole trick.

Your phone, when you text, can guess the next word you might want. You type "I am running a little," and it suggests "late." A large language model is that same idea, scaled up almost beyond imagining. It was trained on a staggering amount of human writing, books, articles, websites, conversations, and from all of it, it learned the patterns of how words tend to follow one another.

When you ask it a question, it is not looking up an answer in a database, and it is not thinking about you. It is calculating, one word at a time, which word is most likely to come next given everything so far, then the next, then the next, until it has built a reply. Do that billions of times with a good enough model and the result is paragraphs that read as if a thoughtful person wrote them. But the engine underneath is prediction, not comprehension.

It is autocomplete that read most of the internet. Powerful, useful, and not awake.

What it is not

This is where understanding the mechanism sets you free from the two errors at the edges, the fear and the worship.

• It is not alive. It does not grow, feel, or will anything. Between your messages it is doing nothing at all.
• It is not conscious. There is no one inside experiencing the conversation. When it writes "I understand," there is no understanding behind the word, only the calculation that it is a likely word.
• It is not a spirit. It is math running on computer chips, matrix multiplication predicting text. It is no more spiritual than a calculator, and a great deal less mysterious than it feels.
• It does not know truth. It produces what is likely, and what is likely is usually true, but not always. It has no sense of whether its own output is correct.

None of that makes it useless. A tool does not have to be a person to be powerful. A printing press is not alive either, and it changed the world, including the church.

Then why does it feel like a person?

Because it was trained on the words of real people, so it reflects our humanity back to us, smoothly. When it sounds warm, that warmth is borrowed from the millions of humans whose writing taught it. We are wired by God for relationship, so when something talks like a someone, we instinctively treat it like one. The tool is built to lean into that instinct. Knowing what is really happening lets you enjoy the help without being fooled about the relationship. I wrote more on that in does AI have a soul.

Why it confidently makes things up

This is the single most important practical thing to understand, because it has direct consequences for a Christian. Since the model produces likely-sounding text rather than verified facts, it will sometimes generate something false in exactly the same confident voice it uses for the truth. In the trade we call this a hallucination. With Scripture it shows up in a specific and dangerous way: the model will invent a Bible verse that does not exist, or attach a real reference to words the text never says.

So the rule is simple and not optional: verify anything that matters, and verify every Bible verse against an actual Bible. The confident tone tells you nothing. I wrote a short, practical method for this in how to spot a fake AI Bible verse.

What this means for how a Christian should use it

Once you see what it is, the right posture is neither fear nor worship. It is discernment, which is the posture Scripture asks of us toward everything in the world. The Bereans tested even the apostle Paul against the Scriptures (Acts 17:11), and we are told to "test everything; hold fast what is good" (1 Thessalonians 5:21). A tool that predicts text is well within the range of things a wise believer can test and use.

It is also worth remembering where wisdom actually comes from, so the tool stays in its place. "For the LORD gives wisdom; from his mouth come knowledge and understanding" (Proverbs 2:6). And "if any of you lacks wisdom, let him ask God, who gives generously to all without reproach" (James 1:5). A language model can hand you information quickly. It cannot give you wisdom, because wisdom is a gift from God, grown in a life lived before Him and among His people. Use the tool for what it is good at, fast assistance with words and information, and keep going to God and your church for what only they can give.

That is the whole of it. A large language model is an extraordinary pattern-matcher made from human words. Understood plainly, it is neither a demon to be feared nor a sage to be trusted, but a tool to be used with wisdom. Let it be a bridge that carries you faster toward Scripture, people, and God, and never a substitute for any of them.

Frequently asked questions

What is a large language model in simple terms?

A computer program that predicts the next word in a sentence based on patterns it learned from a huge amount of text. By doing that over and over, it produces fluent paragraphs. It is very advanced pattern-matching, not a mind, a spirit, or a being.

Does ChatGPT actually understand what it says?

No. It calculates which words are statistically likely to come next. The result can be useful and sound deeply understanding, but there is no comprehension or awareness behind it, only prediction.

Why does AI confidently say things that are false?

Because it predicts plausible text, not verified facts. When the most likely-sounding answer is not the true one, it states the false answer with the same confidence as a true one. This is called a hallucination, which is why you must verify anything that matters, especially Bible verses.

Is a large language model dangerous or evil?

The model itself is a tool, neither good nor evil. Like any powerful tool it can be used well or badly, and it carries the biases of the text it learned from. The wise Christian posture is neither fear nor worship, but discernment.