2) you can at any state check the difference between the simulated state and historical record
3) if your simulation converges, that is the closer to present day, the less the uncertainty, then you are on a right track.
4) if you simulation diverges you go to step 1 and tweak the initial conditions
5) you don't run a single simulation - you run many in parallel and select the best match at each step.
In "Consciousness Explained" Dennet postulates that a conscious mind perceiving a thing-in-the-world need not have perfect rich input data but merely a neuron that responds yes to the question: Am I seeing that thing ?
[1] Consciousness Explained, Daniel Dennet, page 8, [ Prelude, Hard Hallucination Problem ]:
"...intervene somewhere along the chain (anywhere on the optic nerve, ...between the eyeball and consciousness), tugging just right on the nerves would produce exactly the chain of events that would be caused by a normal veridical perception of something, ...producing, at the receiving end in the mind, exactly the effect of such a conscious perception."