Simplification of modern computer architecture…

Say a computer has to calculate the speed / acceleration of a bullet versus a moving target. We can represent this as F = ma intersects a linear locus = velocity x elapsed time.

Say the details aren’t important, so we use the following logic, 

at t1, a is 0, locus is 0

at t2, a is F/m, locus is v.t2

etc.

where they intersect, hit or miss, is v.tN’ and a.tN.tN’. where location of both colliding objects is n1 and n2.

The computer, would take t, n, and v, a as symbols. The processing is simple as t is a running total, n is a fv, and a is v.t.t’ where v is constant.

Such a machine need only consider t++, and the multiplication of v and t in the same clock.

But this gets interesting when we are faced with multiple projectiles as is common nowadays in games. This includes the modeling of a sprite running and firing bullets.

We can see that we need only implement fv over the whole game scene. Same with weather forecasts, but with a few more variables, namely, Work = energy.t , and it is incorporated as such that we have the following variables, force/t = mass.t while F is also equal to ma. We need only add F/t and m.t into our processing. These are simply vectors which alter fv’s vectors.

Say there is reflection, mirror effect, then we need to take fv and complicate it with n/n’. So far so good, minimal waste.

Say the movement is fast. We can use the case, locus towards 0 instead of upping FPS. The direct opposite. To implement the silicon, we need a dedicated fv co-processor or GPU with circa 4 cores or more depending on variables. 

To train AI, it is the same but with a co-processor which matches fv to other variables, the vector being a linguistic, photo, scientific (physics), or logic problem. To achieve processing, the subject is reverse taken apart (reverse engineered) by fv, then compared to another, fv‘ and so on. There is no need for specialized software and hardware.