First let me state that I am not, nor ever have I been, a physics student. I am working on an idea for a book I'm writing.
This is a thought experiment that posits the existence of a computer system that, to cut right to the point, has within its memory banks the thoughts, memories, and experiences of every human being who has ever died on Earth, and this data is continuously updated in real time. Also, this computer system has a constant stream of data fed to it from a network of planetary sensors that record every physical event that occurs on the planet, from the shifting of tectonic plates beneath the Earth's crust to the ol' hackneyed flapping of a butterfly's wings. This computer system is able to process this data through a realizable causal filter that allows it to create an unimaginably detailed model of, well, everything.
This computer system should therefore be asymptotically omniscient, in regard to past and ever-present events that occurred and are occurring. Its data would not be 100% reliable, as there is a distinct difference between observed past (which is subject to interpretation), and actual past (which is indelible), and keep in mind that the physical data fed to it by the planetary sensors would also be considered as observations. I feel that the sheer amount of raw data available would make it so that, when looked at as a whole, multitudinous patterns of causation would emerge.
My question is this: In the example that I described, is there anything that would physically prevent the computer system from anticipating (within specific parameters) future events to a given degree of probability, the accuracy of which would be proportional to the magnitude of the event, and inversely proportional to the time interval between the predicted event and the present?
Furthermore, would it be conceivable that the computer system, when analysing a given prediction, could indicate (within a degree of probability) specific critical points in time and space that, if interfered with by the computer system's operator, could drastically raise or lower the odds of the original predicted event actually occurring?
Further still, the computer system should also be able to simultaneously calculate the potential effects of such an interference, and suggest a course of action that is least likely to create unwelcome consequences further down the road.
Assumptions: The computer system possesses near-infinite processing power and internal storage.
The computer system is programmed to constantly scan for future catastrophic events within a specific magnitude, and when such an event is detected, the computer system is programmed to analyse the various patterns of causation leading up to the event, and suggest optimal points of interference, with the goal of averting the catastrophe.
The computer system operates within a stochastic universe, thus 100% accurate predictions are asymptotically impossible.