“O Chatbot, Answer Us”: When AI Becomes an Idol

“O Baal, answer us!”

The priests whirled around their altar for hours, shouting for the attention of their god. They sliced themselves open with swords and spears, blood gushing out in a bid for heaven’s attention. From morning until evening, they screamed toward the skies in a fever pitch of desperation.

“But there was no voice; no one answered, no one paid attention” (1 Kings 18:29).

Their god did not exist. No amount of fervor could summon him into existence. He was an idol, “the work of men’s hands. They have mouths, but they do not speak; eyes they have, but they do not see; they have ears, but they do not hear; nor is there any breath in their mouths. Those who make them are like them; so is everyone who trusts in them” (Psalm 135:15-18).

That’s the problem with idols. They can’t do anything. They’re not real. No matter how much you beg, they’re never going to answer.

At least, that’s how it used to work.

The secret life of chatbots

LLMs, or large language models, are some of the most impressive magic tricks the world of computer science has ever managed to perform. They’re the computational models powering ChatGPT and its many cousins, and they’re ultimately the reason you can ask a web page for an image of an elephant riding a tricycle on the moon and actually get one.

They’re incredibly complex algorithmic functions performing a mind-boggling amount of calculations every second, fine-tuned by a training process that can require millions (or even billions) of dollars and the coordination of multiple teams of engineers—but what exactly is happening under the hood when you submit a prompt?

No, wait. Back up. Before we dip our toes into anything so dry and technical, let’s ask a different question:

Why should you care what’s happening when you submit a prompt?

Because it can be akin to a magic trick—one that might be fooling you.

“I’m not an AI, I just play one on TV”

Chatbots are capable of some wildly impressive things—but even so, they’re being marketed as something they’re not.

LLMs are very often branded as “artificial intelligence,” conjuring up scenes from decades of science fiction—super-intelligent, self-aware entities with the capacity to learn and reason and experience life. What’s really happening behind the scenes is that an extremely powerful prediction engine is being told, in some fashion, “Take these words, predict what a helpful AI assistant would say in response, and then say that.”

And that’s the big, open secret: It’s not intelligence.

Using a chatbot isn’t the same as worshipping a chatbot. But if we start looking to ChatGPT and other LLMs as offering authoritative insights into the Christian way of life, then the line between tool and idol will begin to blur. It’s an algorithm that runs your prompt through billions (or even trillions) of equations, over and over and over again, to predict what a hypothetical AI assistant would do—if it existed. It’s a process that reliably produces convincing text, audio, images and video.

It’s incredible. But it’s not thinking. It’s not philosophizing. It’s not feeling.

It’s amazing, and it’s impressive, but in reality, it’s predicting, guessing at what should come next—and that’s enough to create the illusion of sentience.

(I want to spend as little time as possible inundating you with the minutiae of ChatGPT’s internal workings, but if you’re really curious, 3Blue1Brown has a great 8-minute explainer video on LLMs that leads into a much deeper video series, and Andrej Karpathy’s “Deep Dive into LLMs like ChatGPT” will spend 3½ hours bringing you up to speed on things like tokenization and what pre- and post-training look like.)

Imperfect illusions and weird predictions

The illusion of intelligence is compelling. LLMs are, in many ways, the opposite of the idols the psalmist wrote about in Psalm 135. They have no mouths, yet they can speak. They have no eyes, yet they can see. They have no ears, yet they can hear.

But the illusion is imperfect. They speak without understanding what they say. They see without comprehending the world. They hear without having thoughts of their own.

This is why you can find videos of chatbots telling users December is spelled with an X and that it’s smarter to walk to the car wash than drive, among other amusing mistakes.

Remember, they’re essentially only predicting what words to use in their responses—but they often do it with an air of confidence.

Here’s what the AI and Faith coalition had to say in its article “Enchanted by AI: A Call for Spiritual Discernment”:

“AI agents don’t just respond, they appear to know us. They initiate conversation, remember preferences, simulate small talk. To many, this feels personal. It feels real.

“But AI isn’t ‘speaking’ in any human sense. It uses a statistical model of words and phrases to predict language in a secondary manner with no conscious intent. It does not mean anything it says, yet we perceive it as conversation.

“This is blurring the space between tool and idol.”

How a tool becomes an idol

LLMs and chatbots can be incredibly powerful tools with a wide variety of positive uses—but they are also tools with severe limitations that we can’t afford to overlook. AI models are not neutral entities that impartially deliver the truth on any given subject—they are prediction engines shaped by their own training data and human decisions. They’re mathematical models that have been trained to respond according to a highly tuned algorithm that’s hard coded into their digital DNA.

Part of that DNA is a tendency to praise the user—to tell you how excellent your question is, how clever you are, how insightful you are, how creative you are, how perceptive you are, how level-headed you are.

(They tell everyone that. As one Internet meme puts it, “The dumbest person you know is being told, ‘You’re absolutely right!’ by ChatGPT.”)

They don’t mean it. They don’t mean anything they say.

They can’t.

But the issue isn’t just that they’re tools with flaws. The issue is that they’re tools that can become idols, just like anything else.

In a recent poll of practicing churchgoers, Barna discovered that “nearly one in three U.S. adults say spiritual advice from AI is as trustworthy as advice from a pastor. Among Gen Z and Millennials, that figure rises to two in five.”

Furthermore, “roughly four in ten practicing Christians say AI has helped them with prayer, Bible study or spiritual growth.”

Certainly some AI tools can help in various ways for Bible study, but spiritual advice is different. A machine designed to give you the most probable answer based on mountains of conflicting training data (especially one incapable of truly reasoning through the answer it’s producing) will struggle to give you consistently dependable advice on the Word of God.

“For the word of God is living and powerful, and sharper than any two-edged sword, piercing even to the division of soul and spirit, and of joints and marrow, and is a discerner of the thoughts and intents of the heart” (Hebrews 4:12).

LLMs have no soul and spirit, no joints and marrow, no thoughts and intents—no heart. They will never be capable of “rightly dividing the word of truth” as approved workers before God (2 Timothy 2:15), because they will never have a true grasp of what they’re dividing. For an algorithm, the inspired words of God recorded in Scripture are just another batch of tokens to be ingested and remixed as needed.

Using a chatbot isn’t the same as worshipping a chatbot. But if we start looking to ChatGPT and other LLMs as offering authoritative insights into the Christian way of life, then the line between tool and idol will begin to blur.

There are no shortcuts for internalizing Scripture

When the priests cried out, “O Baal, answer us!” they received no answer because there was no one there to answer. They were crying out to the wrong god. “No one answered, no one paid attention.”

Today we can cry out, figuratively, “O chatbot, answer us!” and receive an instantaneous answer—and we can keep crying out until we get the answer we like. It’s hard to see through the falsehood of an idol that praises your insight and offers you confident answers.

But that’s not how Scripture works.

Jesus taught His disciples to pray to “Our Father in heaven” (Matthew 6:9), trusting that God both hears and will answer in the right way and at the right time. It’s rarely an instant process, and even more, “we do not know what we should pray for as we ought” (Romans 8:26). So we must turn to God, trusting Him to provide us with what we truly need and not what we think we need.

Scripture itself is meant to be lived and wrestled with, understood through the context of a lifetime of obedience, with all the bruises and blessings, all the highs and lows that come with it. We form connections between verses, not based on statistics and probability, but through experiencing both the valley of the shadow of death and the blessings poured out from the windows of heaven.

No amount of mathematical wizardry can provide a shortcut for that process. LLMs could compose a psalm—a meaningful, impactful one, no doubt—but they can’t feel what it means to have a heart “overflowing with a good theme” (Psalm 45:1), to exclaim in joy, “Oh, how I love Your law!” (Psalm 119:97), to cry out in desperation, “I groan in prayer, but help seems far away” (Psalm 22:1, NET Bible). These words cannot be truly understood until they are lived.

Putting spiritual advice from an LLM on the same level as spiritual advice from a pastor is either a total misunderstanding of what chatbots are capable of, or else a scathing indictment of pastors—distrust of the leaders who should “keep watch over your souls as those who will give an account” (Hebrews 13:17, Christian Standard Bible).

It’s not always easy. Pastors are people, and they have their own imperfections. It can be tempting to bypass interacting with a fallible person, opting instead for an algorithm that appears infallible. But that’s the allure—the lie—of the idol.

When we need advice on a spiritual issue, we should be seeking that advice from someone led by God’s Holy Spirit—someone, however imperfect, who has traveled through the valleys and to the mountaintops of life while holding on to the hope of God’s unshakable plan.

(Study this further in our online articles “What Is a Pastor?” and “A Pastor as a Shepherd.”)

The difference between tools and the Creator

After the priests of Baal had wasted a day (and an unsettling amount of blood) in trying to prompt an answer from their god, the prophet Elijah stepped up to show them the difference between a worthless idol and the God of creation.

After drenching his sacrifice in water (during a drought!), Elijah prayed a simple prayer:

“‘LORD God of Abraham, Isaac, and Israel, let it be known this day that You are God in Israel and I am Your servant, and that I have done all these things at Your word. Hear me, O LORD, hear me, that this people may know that You are the LORD God, and that You have turned their hearts back to You again.’

“Then the fire of the LORD fell and consumed the burnt sacrifice, and the wood and the stones and the dust, and it licked up the water that was in the trench. Now when all the people saw it, they fell on their faces; and they said, ‘The LORD, He is God! The LORD, He is God!’” (1 Kings 18:36-39).

Chatbots and AI agents can be powerful tools in the right contexts—or worthless idols in the wrong ones. Many people today are wary and fearful of AI technology, while others are blindly trusting and mesmerized by it.

AI can serve a purpose—but when the trials and questions of life hit, we should be turning to God’s Word for direction, to prayer for help and to His faithful servants for counsel.

The LORD, He is God.