The prospects are great - a complete replacement of a person, but the current technologies practically do not know how to derive new information from the old, and if they can, then often with errors. Therefore, one can only dream about the automatic construction of any complex logically connected models.
But the future inspires optimism. Three classes of AI tasks, current systems are already quite good at solving: 1) unsupervised classification - to divide data into classes, not knowing anything about the internal data structure 2) supervised classification - to divide data into classes similar to the original the examples given to us. 3) value prediction - to predict the value in a given cell of the table, simply focusing on adjacent rows and columns.
The classes and values obtained in this way on unlabeled data are already quite knowledge.
For more complex systems, only one thing is lacking - greater accuracy. (And for greater accuracy, the main thing is lacking in the speed of computations of modern computers, it is not trivial.)
If you set yourself the goal of looking for a capacious definition for knowledge engineering, it turns out that there are a huge number of those - of varying degrees of capacity. They will differ mainly in the breadth of the definition of the class of problems that can be solved using "knowledge engineering": somewhere you will find narrow formulations like "the problem of compiling knowledge bases for expert systems", and somewhere generous "branch of the development of artificial intelligence. . ". One of the definitions I came across says the following: " Knowledge engineering is the task of gathering and inputting information for use in knowledge-based computer systems. These systems can solve problems or answer questions without the help of a human expert ". I will not undertake to judge the prospects for the future on a global scale, but today it is quite difficult to imagine the functioning of a computer system that recreates some informational model of the world and solves" intellectual "tasks on its basis, without this very "gathering and inputting information". One way or another, some initial data (for example, in the ontological representation) for this model still have to be laid down today with the help of an expert, even for what you mean by AI. By the way, what do you mean by that? ;)