It sounds like deductive logic to me. What is the difference between inferring and deducing? Not sure that I get it.

For example, an LLM may hallucinate the historical fact: “The Treaty of Versailles was signed in 1945 between Germany and France after the second world war” because it sounds reasonable. But armed with inferential understanding, it could realise that “Treaty of Versaille” was after the first world war and 1918, not the second world war and 1945.

Knowledge systems that we’ve had for decades could do that. Prolog can do that. The difficulty is in how to marry the different approaches with deep learning models.