I’ve heard Robert Wright mention on occasion the idea that we are most likely living in a simulation. I’ve only heard the idea in passing but as far as I can tell it goes like this:

At some point in time it will become possible to run a full scale simulation of the our region of the universe.  After that time many simulations will be created. As a result most of the people who exist will exist in simulations. Thus, odds are we are in a simulation.

Now again, I haven’t read the formal description of this argument but the problem for me was simply: why would someone want to construct a bunch of very detailed simulations of the past?

That is, since we can’t now create such simulations we are imagining that in the “real world” it is much later than it is now. What then is the reason for lots of careful simulation of history? Maybe one I could see, but then that gives us at most a 50-50 chance of living in a simulation.

However, I’ve also been reading Eliezer Yudkowsky. So here is my far fetched reasoning for why some more ways we could be in a simulation.

At some point in time someone will create a self improving artificial intelligence, here after known simply as AI.

Once, this AI is created it will become incredibly powerful and its “desires” will determine the future of humanity. As a result it is necessary that the first AI be Friendly. That is, that the AI implement a world that is desirable for humans. Note that it isn’t enough for the AI not to be “mean.” It needs to actively try to create situations that are good for us because otherwise the AI is likely to hurt us by accident.

So far this is standard Yudkowsky.

Here is where I come in. In order for the AI to do this it has to know what the ALL the consequences of its actions are. That means that in some sense it must simulate the future it will produce.

Moreover, because it is important that the very first AI be friendly, it is likely that the people working to develop it would not tell anyone about it. The creators would be afraid that some other untrustworthy group would use the knowledge to create un-Friendly AI.

So, perhaps someone has already created AI in the past. In which case we are the future of the AI. The AI needs to simulate its future and so is simulating us.