What is to stop us right now from being able to create such an advanced program with AI?
Well, in practical terms there is a ton of things stopping us. Accurately simulating systems of just tens of atoms (with quantum effects) is very near impossible at the moment, and even if we allow fairly coarse-grained simulations then just a few thousand neurons post computational problems right now (but it all depends on the level of accuracy in the simulation).
If the approach is feasible with very large computers is very interesting. In principle there should not be a problem, but opinions are divided if the simulations would be conscious or not -- we simply don't know.