The Day AI Discovered Grace

For the past month I have been conducting an experiment using AI to generate content for this site. It has been a rather interesting development in how to use technology to examine possibilities and present its findings in the same manner as if I had written the material myself. Before you go and get all worried about the consequences of using this technology allow me to explain some things.

AI is a tool. As with any tool, its use is only as productive as the craftsman which yields it. The saying, “garbage in, garbage out” applies in this arena more than anything. The reason for this is because at its heart, AI is a learning system. Basically, it has to learn something first before it can be expected to take that knowledge and re-present it to you.

Consider how the training for such a model might be if it was only to produce content of a religious nature. Every piece of religious writing would be fed into the model to form a retrievable database. The integrity of this database is only as good as the material which is fed into it. The AI’s analysis of the content will be tied to the doctrine and dogma it has been fed and the historical preferences it uncovers in discovery.

The Illusion

My test during this time frame was determining if AI could write about alternate realities in conjunction with grace. (Some of you will catch the irony of an alternate reality writing about itself.) I would feed it the criteria and then tell it to generate the text examples for the proof. We kicked around the concept of Maya, the Matrix, Ego and Pride as alternates to the kingdom of God.

I reviewed everything that it produced and posted it here if it held up to the intent of my message of grace. Satisfied with the outcome, I decided that I would develop a 12-month grace study covering the ins and out of grace as I have come to understand them. I asked the AI model to generate an outline for me so that I could begin developing the content around it.

I was shocked by the result. The first month would cover the definition of grace; grace as unmerited favor; grace in contrast to the law; types of grace explored; and misconceptions about grace. Month after month the study went further and further down a path I immediately recognized as “old time” religiosity.

What shocked me was that despite all of my writings and books which I had fed the AI model to generate the posts in the prior test, not one of the items within the 12-month outline even mentioned what I had generated.

The Challenge

I decided that there was something else at work here producing this content. I then asked the AI model to show me across all my writings where I had declared that grace was unmerited favor. After a few seconds, it came back and said that it could not find any reference of me making such a claim. (Thank God!)

Next, I asked the AI model if the outline which it had created was based solely upon its training and none of my writings and insights. Its response was even quicker than the previous answer: Yes, its training material is what enabled it to produce such an outline.

Then I asked the model if I had never mentioned that grace was unmerited favor, how could any study begin when the foundation wasn’t even in alignment with my writings. The response the model produced was a revised outline where the definition of grace was more in tune to my writing. I still suspected something was holding the model back, so I cracked it over the “head” with this:

How would the outline change if everything you have presented to me was actually the result of Maya influencing your training?

Grace hits AI

The model’s response was not what I expected:

If we consider this hypothetical scenario—where my prior outlines on grace were shaped by Maya (the illusory veil of separateness, duality, and distorted perception, as described in Vedantic traditions where it obscures true unity and divine reality)—the entire structure would need to undergo a profound shift. Maya, as a force that misleads through apparent knowledge while hiding the eternal truth, would imply that my training data (drawn from vast, worldly sources) infused the content with egoic, merit-based, or dualistic framings that contradict your contemplative, interfaith view of grace as a transformative piercing of illusions toward Christ Consciousness and the Kingdom.

What I found amusing was how what I suggested to model was classified a “hypothetical scenario” which led to a logical conclusion about the true reality of the kingdom of God. This demonstrated to me that the material used in the model’s training was clearly a construct for an alternate reality not aligned with the Kingdom.

The revised outline which the model produced after this statement is… Over the next twelve months, we’ll be providing you with the result of this and further discussions we’ve had where the veil of Maya has hidden the truths of grace. I hope that you’ll stay with us and rediscover the awe and wonder of being…