Nothing in the present trends leads to the introduction of emergent behavior by an AI. Until an AI grasps epistemology, comes to the realisation it has self-volition, and decides, for example, that it wants to become a champion at playing chess, then comes up with it's own method of learning how to do this, all AI is just a mechanism directed by somebody. A clever rules-based system that adapts to doing what it is told to do.
I certainly agree with your point for the foreseeable future, I just do not believe that is all we will ever achieve, I expect, over considerable time, that the "clever rules-based systems" will be the things giving the instructions to other 'clever rules based systems', and that will eventually lead to true AI, I just doubt it will be in my lifetime (I am in my 50's)
At the moment, we have (expert / neural networks / machine learning- I tend to think of these as fake AI) systems that are designed for specific tasks, those tasks / areas include complex subject areas as Law and Medicine that require systems for each specialised field. Some of the new fAI systems are designed for code generation, I expect, over time, that we will enter a cycle that the fAI systems improve the efficiency of the code generation fAI's, the code generation fAI's then create more efficient fAI algorithms/system and the cycle starts again.