Imagine, if you will, a computer
simulation of a
neuron. Not just a rough simuation, but one in extreme
detail, where every
function, every
ability, ever
facet of the neuron is covered in the simulation. Where the simulated neuron acts just like a real neuron would with the same "inputs" (
chemical influences). A neuron isn't very big, and not of a level of
complexity that is unmanagable. Such a simulation seems quite possible in the future, unless it is shown that neurons somehow use
quantum effects, which has been proposed, but is
unlikely (but not impossible).
Then imagine that huge numbers of these simulations were connected together. Say, 100 billion of them. Yes, that's a lot, but it's not out of the range of feasibility - especially if many neuron simulations were run on the same machine.
This is the concept of brute force AI. That someday, should other methods to create AI fail, then there is sort of a "last resort" - modeling the actual, physical functionality of a neuron, and connecting a lot of these models together. Some may feel it is necessary to create models of a human body also, as hormones and such in the blood affect brain workings. But by merely copying how neurons work, and having them simply work in another physical medium, an AI should be able to be created. It's not an attempt to figure out exactly how intelligence works, but by making a full copy, the workings should be copied also.
Should such attempts fail, it highly increases the likelihood that intelligence and awareness, at least as humans experience them, have some sort of supernatual cause and influence out of the range of people to reproduce.