God in a Box --> The Simulation Problem

We recreate systems inside virtual worlds to better predict and understand them. This is simulation. The more granular/high resolution the simulation, the more accurate* it's results (usually).

If you want to understand/predict intelligent life, you simulate it. If you want to simulate it well, you do so with high granularity. The higher the granularity, the more the simulated life is actually alive. This is the traditional (ethical) simulation problem.

If you want to predict whether creating a greater intelligence would be safe, you simulate it and see how it acts/ if it escapes in the simulated world. The problem is that you likely cannot keep a god in a box. The super-intelligence may be able to realise it is simulated and escape the simulation. This is the AI simulation problem?

* Need to introduce probability/absolute vs general dsitinction into english