I have experience with large NLP models and here are my humble thoughts: In a large language model you can refine your original request via conversation. In a

I have experience with large NLP models and here are my humble thoughts: In a large language model you can refine your original request via conversation. In a search engine you tend to rewrite your query to refine it. LLM conversation is much more stateful than an almost stateless query to a search engine. However a search engine shows you more than one answer. With LLMs you can't easily add knowledge graph style facts (subject predicate object). I has either memorized a fact from the training set or not and you can't easily insert one more fact in there. retaining LLMs is not cheap by the way. An ideal search engine for me is conversational, up to date and one that gives you multiple answers.