Why Your Relationship With Him Works Like the Most Powerful Technology Ever Built
By Timothy · · 10 min read
Most people understand machine learning as the technology behind ChatGPT, recommendation algorithms, and self-driving cars. What most people don't realize is that the same foundational principles powering the most sophisticated AI systems ever built are also the oldest principles of spiritual formation ever recorded.
This isn't coincidence. And it isn't metaphor for its own sake.
Before the analogy, a brief and accurate description of the real thing.
Machine learning is a branch of artificial intelligence in which a system learns from data rather than being explicitly programmed with rules. Instead of writing code that says "if this, then that," you feed the system enormous amounts of data and let it discover patterns on its own. Over time, through repeated exposure and continuous error correction, the model develops the ability to recognize patterns, make predictions, and generate outputs that no human explicitly wrote.
A machine learning model becomes what it is trained on. Its intelligence, its accuracy, its capabilities, and its blind spots are all a direct function of what it has been exposed to.
Change the training data and you get a different model. Change the frequency of training and you get a less capable model. Remove error correction entirely and you get a model that confidently produces wrong answers.
Now hold that framework, and open Genesis 1.
Here is the uncomfortable truth this analogy reveals immediately: whether you are intentional about it or not, your mind is running a training process right now.
Every piece of content you consume is training data. Every conversation you have is training data. Every social media feed, every news cycle, every environment you sit inside is feeding inputs into a model that is continuously being refined — and that model is you.
Do not conform to the pattern of this world, but be transformed by the renewing of your mind.
Romans 12:2
The pattern of the world is not neutral. It is an active training program with a specific output in view — one that makes you increasingly shaped by anxiety, comparison, consumption, and performance. The question is not whether your mind is being trained. It is whether you are choosing the training data, or letting someone else choose it for you.
In machine learning, there is a foundational principle that engineers learn quickly and never forget: garbage in, garbage out.
A model is only as good as the data it was trained on. Feed it biased data and it produces biased outputs. Feed it noisy data and it produces unreliable outputs. Feed it incomplete data and it develops blind spots it cannot even detect in itself. You can have the most advanced neural network ever built and still produce a worse model than a simpler one trained on better data.
These commandments that I give you today are to be on your hearts. Impress them on your children. Talk about them when you sit at home and when you walk along the road, when you lie down and when you get up.
Deuteronomy 6:6-7
That is a training data specification. It is not describing a once-a-week lecture. It is describing saturation — every context, every conversation, every transition of the day becomes an opportunity for input. The goal is not information transfer. The goal is pattern formation.
Philippians 4:8 gives the same instruction in list form: whatever is true, noble, right, pure, lovely, admirable — think about such things. This is a curated dataset. Paul is telling you which inputs to prioritize and which to deprioritize. This is not censorship. It is data hygiene — the same discipline every serious machine learning engineer applies before they ever start a training run.
One of the most consistently confirmed findings in machine learning research is that frequent, incremental training outperforms rare, intensive training.
A model trained on small batches of data repeatedly, with continuous feedback and correction, develops far more robust and generalizable capabilities than a model trained once on a massive dataset with no further refinement. The brain works the same way. Neuroscience has confirmed what scripture declared: repetition builds neural pathways that become the default routes your mind travels.
This is why a single powerful sermon rarely changes a life permanently. And it is why daily exposure to scripture, over years, produces people whose thinking is qualitatively different from those who don't. The change is not dramatic. It is the accumulation of training epochs — each quiet morning, each returned-to passage, each practiced response — building something that occasional intensity cannot replicate.
But solid food is for the mature, who by constant use have trained themselves to distinguish good from evil.
Hebrews 5:14
Constant use. Trained themselves. The Greek word translated "trained" is gymnazo — the same root as gymnasium. It implies disciplined, repeated exercise with a developmental goal in mind.
In machine learning, the loss function is the mechanism that tells a model how wrong its predictions are. After each training pass, the model's output is compared to the correct answer, the error is measured, and the model's internal parameters are adjusted to reduce that error on the next pass. Without a loss function, the model has no way to know it is wrong, and no mechanism to improve.
The Holy Spirit is the loss function of the spiritual life.
John 16:8 says the Holy Spirit will convict the world concerning sin and righteousness and judgment. Conviction is not condemnation — it is correction. It is the mechanism by which you learn that your output diverged from the target. Without it, you can be completely confident in outputs that are completely wrong, like a model that has never been corrected and has no idea it is predicting backwards.
The willingness to be corrected is not weakness. It is the most important variable in whether training produces growth or produces stagnation. A person who resists conviction, who dismisses the discomfort of correction, who curates only feedback that confirms what they already believe — stops growing. They may become more confident. They do not become more accurate.
Whoever loves discipline loves knowledge, but whoever hates correction is stupid.
Proverbs 12:1
The process by which a machine learning model improves is called gradient descent. The model measures the error, calculates the direction of steepest descent toward the lowest error, and takes a small step in that direction. Then it measures again. Then it steps again. Thousands of times. Millions of times. Each individual step is almost invisible. The accumulated result of all those steps is a model that can do things it could not do at the beginning of training.
Nobody looks at gradient descent and says "this step didn't matter." Every step matters. The direction matters on every pass, even when the movement is imperceptible.
And we all, who with unveiled faces contemplate the Lord's glory, are being transformed into his image with ever-increasing glory, which comes from the Lord, who is the Spirit.
2 Corinthians 3:18
Being transformed. Present continuous tense. Not was transformed once. Not will be transformed eventually. Is being transformed — right now, step by step, epoch by epoch — into something that looks more like the target than it did at the start of training.
One of the most instructive failure modes in machine learning is overfitting. This occurs when a model is trained so heavily on a narrow dataset that it performs perfectly on that specific data but fails completely when exposed to anything outside of it. The model has memorized the training examples instead of learning the underlying principles. It cannot generalize. Real-world complexity breaks it.
This is legalism.
A faith that is optimized for a specific set of rules, a specific cultural expression, a specific set of acceptable behaviors — and then encounters a person or situation outside those parameters — discovers that its model doesn't generalize. It becomes brittle. It produces wrong answers with high confidence. It is technically well-trained and practically useless for the situations that actually require wisdom.
Jesus confronted overfitting continuously. The Pharisees had trained their models intensively on the law — and their loss function had become approval from other Pharisees rather than alignment with God. They had overfit to a narrow dataset and lost the ability to recognize the very thing the whole training was pointing toward when it showed up in front of them.
The solution to overfitting in machine learning is regularization — techniques that force the model to learn general principles rather than memorize specific examples. Grace is regularization. It prevents faith from becoming so rigid and narrow that it can only function in controlled environments, and forces it to engage with the full complexity of people and situations as they actually are.
In machine learning, there is a distinction between training and inference. Training is what happens during development — the learning process. Inference is what happens when the trained model is deployed in the world — making real predictions on real inputs, in real time.
All the training is for the inference. A model that is perpetually training but never deployed is not actually useful. The goal was always the output.
The goal of a relationship with God, of consistent study, of daily prayer, of ongoing correction and refinement — the goal is inference. It is the moment when someone asks you a hard question and you have the answer. When a situation requires wisdom and you have it. When you encounter someone in pain and you know what to do. When a door opens and you recognize it immediately because your model has been trained to see it.
A good man brings good things out of the good stored up in his heart, and an evil man brings evil things out of the evil stored up in his heart. For the mouth speaks what the heart is full of.
Luke 6:45
What comes out is a direct function of what went in. The inference is only as good as the training.
The machine learning framework is not new theology. It is a new language for something scripture has been saying since Moses stood before Israel and described the training program God had in mind.
Love the Lord your God with all your heart and with all your soul and with all your strength. These commandments are to be on your hearts.
Deuteronomy 6:5-6
That is the model objective — love God completely — and the training methodology: continuous, saturating, deeply embedded exposure to His word in every context of daily life.
The smartest technology humans have ever built turns out to operate on the same principles God gave to Moses in the desert.
Which means that when you sit down with your Bible on a Tuesday morning that doesn't feel spiritual, you are not just doing a religious ritual. You are running a training epoch. The model is updating. The parameters are adjusting. The output is being shaped, incrementally, toward something.
Show up enough times. The model changes. That is not an inspirational statement. It is machine learning. And it was always the point.