WHY AI WILL NEVER SURPASS HUMAN INTELLIGENCE

A paper entitled “The unreasonable effectiveness of deep learning in artificial intelligence” argues that the way forward towards achieving general AI, that is to say a human level intelligence, is to copy how an organic brain does if for humans. The paper argues that AI has to move from a very limited 2D-space which is referred to as Flatland to a tera-dimensional space that represents the million billion synapses between the neurons in the cortex of the human brain. It is pointed out that the move from 2-D AI to tera-dimensional AI is actually a move in the wrong direction if they ever hope to achieve general AI. The fact is that although there are a million billion synapses between the neurons in the cortex, human consciousness is one dimensional or holistic. In order to achieve general AI the machine will have to do everything a human can do where there are no gaps or seams in the output. A model of the human brain is offered where different sections of the cortex are specialized for different functions and these disparate regions communicate with each other electronically at the speed of light via brainwaves and this is how the brain generates a global holistic 1-dimensional consciousness in us. Also as numbers don’t exist in Nature an organic brain, unlike deep learning, generates intelligent output without the aid of numerical programs or statistics.

Download PDF file