There is apparently a generational divide in the use of AI. Those of us who witnessed the transition to the ubiquity of verbal machines tend to use it for search, while those new to the thing just dialogue with it and get it to say things. Arguably, the people who understand the nature of the machine have a point.
It's as if there's an immense lawn on which you know there is a single leaf, despite the fact that you've just raked the whole thing. Now, you could go back and rake the whole thing, or you could use your knowledge of where the trees are, and where such a thing might tend to be, or you could call up to the aeronaut hovering in his Montgolfier balloon above the estate, an he can tell you where the leaf is. Now, the utility of such a fellow to me is that he knows where that precise bit of information is, but his actual function is more katascopic, in that he sees the whole lawn. What it does is fundamentally different than what I want from it.
Now, someone who hadn't raked the lawn might not know where the trees are, or where the leaves were likely to be, but they could similarly pick up a pair of semaphore flags and make the same inquiry of the balloonist. In the same way that people who witnessed the transition tend to keep their files in logically structured folders, while the newcomers just use search to find what they need from the fungible mass.
If you only want to know where the single leaf is, the machine will give you an answer. But the logic that surrounds the leaf is no longer the knowledge of the tree and the land. The katascope works by negotiating the logical structure of language, predicting the next word recursively, using the same logic that created the corpus of language that it is ransacking. If you know nothing of the tree and the land, you'll get the full utility of all past words spoken on the matter, and the logical structure of their relationships. If, however, the logic within you is what gives that bit of information meaning, your encounter with the machine presents a logic that differs slightly from the corpus.
But if you just want to think as everyone thinks, it might be a rather useful way of calibrating your logic. Perhaps the generational divide has less to do with different perspectives on the machine, and more to to with people with different degrees of wisdom, which is to say, different relationships to the accumulated logical structures of the past.