Well, if you consider a neural network that has a depth of d layers and a width of n neurones on each layer, then you will be able to:

• 1/ implement a deep computation with a network that has d >> n,
• 2/ simulates lots of possibilities with a network that has n >> d

Where:

• “>>” means “very greater than”,
• “deep computation” means that the algorithm mimicked by the neuronal network is expected to have a lot of steps involved,
• “simulating lots of possibilities” means that you’d try to evaluate possible futures, and that is the purpose of intuition.

—

So, 1/ is CPU-like and 2/ is GPU-like, they are serial processing units and parallel processing unit, and generally speaking in any brain, we need both:

• 1/ the serial processing unit (SPU) would be for handling memory and representation of the world, witch is the past,
• 2/ the parallel processing unit (PPU) would be for anticipating the future.

—

And of course, the best for doing so would be to use different technologies for those 2 goals:

• whereas discreet computations based on natural numbers would suit best for SPU,
• a more appropriate suit for PPU would be vectors, or waves (with quantum computing)

And of course, vectors are already what’s used on GPUs. So the next step is to use quantum computing, and there you’ll be able to have great Intuition’s AIs.