Why deep learning image recognition alone won't generate artificial intelligence and what will

Deep learning based image recognition (of cat for example) in an image is a major accomplishment for AI but that classifier won't create intelligence like a human. It may generate a pattern recognizer that equals or surpasses human ability, but not something like human intelligence. If we want a machine to learn intelligence, we probably need to make the system learn in steps similar the the way human evolved. First create a system that has an agenda (we evolved that as worms looking for food). Then, incrementally evolve a system for motion (limb control to find food). Then evolve abilities to reason and interact with the environment to accomplish the agenda more effectively. Then evolve advanced communication skills. Each one of these is a complex learning task that evolution optimized through genetics, but we would want to optimize them through some learning process like simulated annealing or some genetic algorithm. The point is that the intelligent system needs to learn first how to have an agenda, not just a recognition task. Then it will begin to be more "alive" and interactive in interesting ways.