Statistical models, according to someone part of artificial intelligence, explain a phenomenon of interest using other phenomena. They try to go from the particular (row of data, observation or sample) to the general. We NEVER have all the data, so as even the inductive turkey who thought he was getting food every day, except becoming food on Thanksgiving, remembers, this approach has limitations.
In contrast, the deductive method goes from the general (or universal, in the case of physics) to the particular. Logical mechanisms are usually encoded through “if/else.” Hence a famous objection to those who develop artificial intelligence: this is trivial computing with conditionals. Expert systems fall into this category, and you can find an article about it. So do certain monitoring and planning systems.
Just as the inductive method has a limitation on data, the deductive method also has a limitation: the validity of what is reputed to be general, e.g., a rule (from which if/else is constructed). In more technical jargon, the assumptions, the axioms.
So since both methods, or intelligences, have limits, does usefulness lie in the middle? No, it lies in the depends and from the consciousness of the consultant.
In fact, as Aristotle also reminds us, intuition facilitates induction and deduction.
One could say that technologies such as GPT, also called generative (even creative) artificial intelligences, come close to deductive intelligences because they have a huge amount of data. Let’s leave aside the discussion of the quality and generality of that data. But one could also say that that huge amount of data allowed them to learn rules. Not just grammatical ones.
But in what tasks is it worthwhile to use technologies like GPT, in your company? Also to reduce the workload of some employees. Let’s discuss this in a free call of about 40 min, where I will first inquire about your business.