prediction and validation——explore into the future

Elon Musk talking about mars. we don't have to actually send men to mars. we could use remote control mode, one main problem is the time delay of the command transmission and the feedback return. 

1. "built a reality": use the virtual reality to copy the "past" conditions, create a current world of the "past".

2. prediction: we give the command to change the "virtual current" -- the 'past'. the command were send out to the remote site. base on the physical model of the current world, we predict the local result before it actually happened. 

3. validation: the command reach the remote site, and change its condition. the result send back to local control. compare the prediction to the feedback -- virtual current, find the different, and fuse the feedback into the virtual reality like patches. so we create a changing virtual reality fusing the past condition and future prediction of the remote site.

to achieve the most like-to-happen, it requires the precise physical model and consideration of the randomness. it seems impossible to do, cause the possibility maybe too many(could be solved by the computer) or the randomness dominate the result(like the quantum effect, it can not be precise predicted). like drop a glass on the ground predict the shatters' shape and location, but on the other hand, the condition could be manually simplified that glass is plastic, and it's just bouncing and rolling, and the final possibility could be relative less and more acceptable(you just don't want step on it, and could wait to pick it up later). 

are our behaviors all strongly chained together? our reactions are all highly depended on the simultaneous assessment? it seems like that, but how much we can tolerant the "not really"?