Simulation or Not

Every few years the question comes back, louder than before. Are we living in a simulation? The Matrix made it pop culture in 1999. Elon Musk gave it a sound bite in 2016 when he put the odds at billions to one that we are in base reality. Philosophers gave it a formal argument in 2003. And every leap in computing power makes it feel a little more plausible.

It deserves better than a shrug. The question is really three questions wearing one coat: Is it possible? Is it likely? And if the answer is yes, does anything change? The honest answers are: probably, maybe, and no. This article is about all three.

In 2003, Oxford philosopher Nick Bostrom published Are You Living in a Computer Simulation? in the Philosophical Quarterly. It is the version of the argument that serious people engage with, and it is not what most people think it is. Bostrom does not claim we are in a simulation. He claims that at least one of three propositions must be true:

• Almost all civilizations at our stage go extinct before becoming technologically mature. We never get there.
• Technologically mature civilizations are possible but almost none of them run ancestor simulations. They could, but they do not.
• We are almost certainly living in a computer simulation.
• The logic: if many civilizations reach maturity and run vast numbers of simulated minds, then simulated minds vastly outnumber real ones, and the odds say you are simulated. It is a statistical argument, not a physics argument. That distinction matters, because it is also where the argument is weakest.

The argument has force because its premises keep getting more plausible. Computing power has grown so fast for so long that simulating a universe no longer sounds like fantasy; it sounds like an engineering problem with a long timeline. Video games went from Pong to photorealistic worlds in fifty years. Extrapolate that curve a thousand years and the idea of a simulated ancestor stops being absurd.

There is also the substrate-independence assumption: if consciousness is a pattern of information processing rather than something unique to biological brains, then a sufficiently detailed simulation would contain conscious minds. Most working AI researchers and neuroscientists lean this way, even if they would not put it in those words.

And there is a quieter point. Every civilization that survives long enough will want to study its own history. Running ancestor simulations is one of the cheapest ways to do it. The motive is not exotic. It is archaeology.

Now the other side, because an argument this big needs its critics heard. The most common objection is that the hypothesis is unfalsifiable: no experiment can distinguish a perfect simulation from base reality, since the simulation defines what your instruments report. Critics like physicist Sabine Hossenfelder argue that an unfalsifiable claim is not science; it is metaphysics with better marketing.

Then there is the physics problem. If the universe is rendered, we might expect to see rendering artifacts: a minimum pixel size, shortcuts in unobserved regions, computational limits. Some point to the Planck length as a candidate pixel. But the universe keeps behaving like it is continuous and uncompressed, and every supposed glitch (cosmic ray cutoffs, quantum weirdness) has a boring conventional explanation. The absence of evidence is not evidence of absence, but twenty years of looking has produced nothing.

There is also the regress problem: if we are simulated, our simulators are probably simulated too, turtles all the way down, and the statistics of the trilemma get murky at every level. And philosopher David Chalmers, in Reality+ (2022), makes the subtlest point of all: even if the simulation hypothesis is true, simulated reality is still real. Virtual experiences are genuine experiences. The world does not become fake; it becomes differently implemented.

This is the part most discussions skip, and it is the only part that matters for how you live. Suppose Bostrom is right and this is all running on some civilization's hardware. What, exactly, would you do differently tomorrow morning?

You would still drink your coffee. You would still keep your promises, because the people you promised are conscious whether they are made of neurons or numbers. You would still build your business, because your customers' problems are real problems inside the only world any of us can access. A perfect simulation is, from the inside, indistinguishable from reality, which means reality is whatever you are inside of.

The moral weight of your choices is identical whether the world is base reality or rendered. That is not a dodge. It is the conclusion the philosophy actually supports: Chalmers' point, carried to its end. Ethics does not need base reality. It needs conscious beings who can suffer and flourish, and those exist in every version of this story.

The simulation question is fun, but here is the forecast people actually need. The next three years will not bring a revelation that we are simulated. They will bring something more concrete: AI systems that act in the world on their own. Agentic AI that books, buys, negotiates, and builds without a human in the loop for every step. That is already starting, and by 2029 it will be ordinary.

For an SMB, that means the competitive line moves. The businesses that survive will be the ones whose information is structured, verified, and machine-readable, because AI agents will be the ones reading it. The ones that stall will be the ones still waiting for a human to click through their homepage. This is not a prediction about consciousness or reality. It is a prediction about plumbing, and plumbing is where the money is.

The honest uncertainty: nobody knows the pace. Timelines in AI have a habit of being both too optimistic and too pessimistic at once. The direction is clear; the speed is not. Plan for fast, build for uncertainty, and keep a human approving the meaningful changes. That last part is not philosophy. It is operations.

Are we in a simulation? The serious argument says it is possible and maybe likely. The honest objections say it is untestable and currently evidence-free. The practical answer says it does not matter, because the world you can touch, the people you can help, and the work you can do are all real in the only sense that counts.

So build like it is real. Be kind like they are real. And keep asking the question, because the asking is what makes the mind worth simulating in the first place.