Reasoning is the disadvantage of ChatGPT, Are there some method to make ChatGPT and Auto-regress model attributable and interpretable? Is "Objective-Driven AI

Reasoning is the disadvantage of ChatGPT, Are there some method to make ChatGPT and Auto-regress model attributable and interpretable? Is "Objective-Driven AI architectures" just a vision? Is there some paper about that?