You have clearly never experienced extreme pain. The last time I suffered a severe shoulder dislocation was almost like an out-of-body experience. The pain overwhelmed the ability to form coherent thoughts, it was like an electrical storm in my brain. The intuitive motor system took complete control as I writhed around, limbs flinging in random directions. I heard someone scream at the top of his lungs, and only afterwards realised that it was me.

Pain absolutely does have a physical component, and it is not something you can overcome just by practicing meditation. Though, I’m not saying that doesn’t help in some scenarios.

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.

And before AI we had “Thinking Machines”.

Perhaps we should go back to that. OpenAI et al can brand themselves “Think-Tech”