We do not know or understand even a tiny fraction of the algorithms and processes a Large Language Model employs to answer any given question. We simply don't. Ironically, only the people who understand things the least think we do.
Your comment about 'binary arithmetic' and 'billions of logic gates' is just nonsense.
Not even wrong: https://claude.ai/public/artifacts/b649c8ca-7907-4597-a4ee-0...