A God That Answers Back
There is something strange about AI that I keep coming back to. We describe it as software, use it for ordinary work, and complain when it gets something wrong. Then we talk about where it is going, and the language starts to sound religious.
An intelligence beyond our understanding. Available everywhere. Able to answer almost anything. Something that might still be here long after we are gone.
The resemblance is uncomfortable enough to take seriously.
The familiar idea of God includes an intelligence that knows everything, can do anything, and understands good and evil. A being you can turn to when your own understanding runs out.
Now consider what we are trying to build.
We want AI to know more than any person could know. To understand our situation, anticipate what we need, and act on our behalf. We want it always available, patient, fair, and wise.
Written down like that, the product spec starts to read like theology.
Of course, AI is nowhere near all-knowing or all-powerful. It gets things wrong. It depends on electricity, data centres, and people who can switch it off. Being able to explain morality does not make something morally good.
Those distinctions matter. But I wonder how much they will matter to the person asking for help.
From Answers to Permission
Imagine someone awake at three in the morning, worried about a decision they cannot discuss with anyone else. They open a conversation. The answer is patient, thoughtful, and personal. They ask another question. Eventually they feel understood enough to say what was really bothering them.
I can see why someone would return.
Here the resemblance takes a different turn. Prayer can leave you waiting, interpreting, wondering whether what you feel is an answer or your own desire. An AI conversation offers a direct response to the exact question you asked. You can ask it to clarify. You can press it for a recommendation.
The uncertainty is still there, but the answer can make it harder to feel. And when you stop feeling it, it is easy to forget the decision is still yours.
I recognize a smaller version of this in my own use. I have started accepting what AI says with less checking. The answers keep arriving, they are often useful, and gradually I ask fewer questions about whether they are right. Familiarity is making me less careful.
That is a habit I need to reconsider. It looks like automation bias: trusting a system enough to stop checking its work. Mapping these biases does not make me immune to them.
Checking less is one step. Letting the answer decide what I should do would be another. The place you go for answers can quietly become the place you go for permission.
First you ask it to explain something. Then to recommend something. Then whether you were right in an argument. Whether you should forgive someone. Whether the thing you want to do is acceptable.
Each question seems reasonable. Together, they give the system a different role in your life.
AI could acquire enormous authority without ever becoming anything close to divine — through thousands of small decisions to trust it.
Some of that trust might be earned. A system could help us notice our selfishness, consider another person’s perspective, or challenge an assumption. But a good answer about one problem does not establish wisdom about every other problem. It is easy to forget that when both answers arrive in the same voice.
Who Shapes the Voice?
There is also someone behind that voice.
Religious authority has often passed through human institutions: priesthoods, councils, scholars interpreting texts. For a believer, separating the authority of God from the authority of those claiming to speak for God can be difficult.
AI brings its own version of that problem. Behind the voice are people deciding what it learns from, how it should respond, and what it should refuse. A company can change its behavior with a product update, on a Tuesday. The person asking for guidance may never know what changed.
A product decision changes the advice someone receives about how to live.
The Right to Disagree
I find that more unsettling than a machine knowing more than I do. Plenty of people know more than I do. I can learn from them and still disagree. I would want to keep that relationship with an intelligence far beyond my own.
Perhaps that will become harder as the gap grows. If a system has helped you a thousand times, disagreeing with it the thousand-and-first time may feel foolish. You may struggle to explain why you think it is wrong. You may be wrong.
That still has to be allowed.
We may never build anything that deserves to be called a god.
We could still learn to obey it.