ok youve got a temporal perceptron feeding in motor and vision, it has a biasing ability to pick the pictures that go with the motor. anyway, if you can pick pictures that go with motor, you can now use the history to look into the future in a million different ways. if certain sense combinations, that are on the surface of the body of the robot, are used an invarient triggers to apply to its eye and motor combinations, now you can form conceptual relations, if certain triggers indicate success in whatever mode its in, basicly get the right order of triggers, and you have created an intelligent organism, relying on instinctual relations alone. it can learn how to kill things if it can make sense of body triggers that tell it it has won. it will act a little like a simple animal, an organism, but if you look into the history with enough computational power it can learn from complete random mutation, abiding by the logic of finding the right triggers, which are being continually growing from the invarient sense responses. this may be how you can get to human like intelligence but you must work out how we invent more triggers, and develop a new order of triggers, and thats all conceptualizing even is. but as a simple organism that can work out how to kill you, its very successful at it. a trigger leaves a trail that connects to it, in the brain rock, so it knows how to find them.
conceptualization is just development of new triggers and behavioural order.
prediction is cunning and learning speed.
it actually comes from learning jeff hawkins htm, and crossing it with karl sims, evolved creatures.
and you get the creature in a single body, and thats the beginning, just now youve got to draw more relations, and youve got a human.