Does God Exist?
How the DeepMind documentary made me think more on this
I recently watched the Javed Akhtar vs Mufti Shamail debate. Two hours, two intelligent minds, one question: Does God exist? First of all, Huge kudos to the organisers and both speakers for reminding us what meaningful debate can look like.
My belief has previously been somewhat similar to Javed Akhtar. His argument: If God is all-powerful and all-merciful, why does evil exist in the world? The response that God created the possibility of evil but isn’t evil himself felt inadequate. Couldn’t an omnipotent being design free will without the bombing of innocent people?
I’m not an atheist. More agnostic, leaning toward “probably no god.” The universe feels too random, too brutal, too indifferent for someone to be running it who actually cares.
Then I watched a documentary about Demis Hassabis and DeepMind. I kept thinking more and more about the question, and got reminded of a third probability and how Google DeepMind is actually doing that.
What DeepMind is Building
DeepMind has built AIs called Genie. Given a single image, say, a photograph of a waterfall in California, it can generate a navigable, game like 3D environment.
Then they drop AI agents called SIMA into these worlds. The agents have goals: navigate this maze, climb that ladder, maximize your score. They fail. They learn. They improve. Endless trial and error.
From a SIMA agent’s perspective, that generated world is reality. They don’t know their creators exist. They don’t know they’re in a simulation. They simply exist, strive, fail, and try again.
The Third Option
The debate offered two familiar positions.
The Mufti argued that God must exist because existence itself cannot regress infinitely. There has to be a first cause, something which wasn’t created. Akhtar countered that even if such a being exists, it cannot be both all-powerful and all-merciful, because a just and caring god would not permit horrors like the suffering of children.
Watching DeepMind create worlds and populate them with goal-seeking agents, I began to think of a third possibility, one that neither side addressed.
Maybe there is a god. Maybe he just doesn’t care about us.
This wasn’t an entirely new idea for me. I remember reading something similar in Yuval Harrari’s book Homo Deus. But seeing DeepMind build worlds and release agents into them gave that intuition a concrete shape.
Do the researchers care about individual SIMA agents? Do they mourn when one fails? Do they intervene when an agent keeps hitting the same wall? No. They’re running experiments. They’re interested in the system, the emergent patterns. Individual agents are data points.
We might not be anyone’s children. We might just be agents.
This answers Akhtar’s Problem of Evil more cleanly than the Mufti’s defense. Why does God allow suffering? Because he’s not watching. He set up the world, dropped us in, and he’s waiting to see what emerges. War isn’t a theological puzzle if the creator is indifferent. It’s just what happens when you let a simulation run.
The Simulation Question
Philosopher Nick Bostrom made a famous argument: if advanced civilizations can create realistic simulations, and they have any interest in doing so, the number of simulated beings would vastly outnumber real ones. Statistically, we’re probably simulated.
Recent calculations put it at roughly 50-50 odds we’re in base reality. But if we ever create conscious simulations ourselves, those odds collapse. We would almost certainly not be real.
The Uncomfortable Parallels
DeepMind creates worlds. So does the god of every religion.
The system gives agents purpose—score points, navigate mazes. Religion says we’re created with purpose too.
The creators are invisible to their creations. The Mufti argued God is known through reason and signs, not physical detection. Same structure.
They allow failure and suffering—that’s how agents learn. The Mufti’s free will argument, essentially.
They can end the simulation whenever they want.
Where This Leaves Me
I’m not suddenly religious. I’m not convinced we’re in a simulation. But my leaning towards their being no creator has changed slightly.
This feels more honest than “mysterious ways.” It makes sense of the world’s cruelty without requiring deep theological thinking.
Somewhere on a DeepMind server, a SIMA agent just hit a wall for the thousandth time. It doesn’t know why. It can’t conceive of Demis Hassabis. It simply tries again. Tomorrow morning, I’ll brew my coffee for the ten thousandth time, commute the same route, solve the same kinds of problems.
Maybe that’s the point. Maybe it isn’t. We keep going anyway.
Author’s note: I used AI as a thinking and editing partner while writing this essay. The ideas, arguments, and conclusions are my own.