Hardwiring the value of those coordinates at time t+s would be incompatible with other parts of the simulation, and ill-defined.
I’m no longer sure this would be such a serious problem, if we had a very good deterministic physics engine devised and built (and thus understood) by ourselves. Yes, the AGI’s appearance would be a discontinuity in the laws of physics, which would never actually happen. But why can’t we just take the whole state of the universe at time t, then rewrite a certain cube of matter by a box containing the AGI, and then compute deterministically the evolution given by the laws of physics from there? I guess my worries were in the vein of “everything is connected, so this will amount to a discontinuity in other parts of the world outside the box, for instance in the gravitational or electromagnetic fields emanated from the box”. But why is this discontinuity a problem? We can just set it up, and keep computing continuously from there.
I’m no longer sure this would be such a serious problem, if we had a very good deterministic physics engine devised and built (and thus understood) by ourselves. Yes, the AGI’s appearance would be a discontinuity in the laws of physics, which would never actually happen. But why can’t we just take the whole state of the universe at time t, then rewrite a certain cube of matter by a box containing the AGI, and then compute deterministically the evolution given by the laws of physics from there? I guess my worries were in the vein of “everything is connected, so this will amount to a discontinuity in other parts of the world outside the box, for instance in the gravitational or electromagnetic fields emanated from the box”. But why is this discontinuity a problem? We can just set it up, and keep computing continuously from there.