The supposed similarities between artificial intelligence and human intelligence have been made often enough. Neural networks have been widely advertised as an attempt to mimic the neurons in the brain, but this is a fairly superficial analysis. In reality AI is already well beyond any attempts to mimic human intelligence simply because it is more productive to try and create intelligences that are based on fundamentally different ideas.
The point has been well made that we made more progress with flying machines when we stopped trying to mimic birds flapping their wings, and instead focused on the science behind flight – on aerodynamics. So it is with AI. The Turing test is an attempt to emulate human behavior, and an AI passes the test if it is impossible to tell the difference between the AI and a human being. But more recent thinking simply asks the question – why try and emulate human beings?
The dominant model for AI right now is the rational agent . An agent is simply something that does something, and a rational agent is given a goal it needs to work toward. The better it scores in this respect the more we say it is rational. You might think this emulates human behavior, but in the main it does not. Human beings find it very difficult to behave in a rational manner, with emotions and conflicting goals creating sub-optimal behavior. A rational agent is single minded in the way it tries to maximize the goal it has been given, with no emotions or conflicting goals to get in the way. So the notion of the rational agent already makes AI different from human beings – but there is more.
An AI is capable of dealing with thousands of inputs – different data sources and/or sensors. No human being can process such a large number of inputs. A robotic device for example might have infra red, visible light, radio wave and other sensors. This will give it a wholly different perception of reality to the one human beings perceive. Very long wavelength electromagnetic waves for example can bend to some extent around corners. We are conditioned and limited by our five senses. A robotic device need have no such limitations, with possibly hundreds of “senses” giving it a wholly different view of reality.
We already have the situation where the astonishing success of deep learning networks is hardly understood. As AI moves into the work place, medicine, entertainment and so on, so we humans will strain to understand their behaviors, and that is of course quite dangerous.
It is a mistake to think that AI will mimic human behavior. It will not. The behavior of AIs will be barely understood by human beings, and that should cause some concern.