This is a very valid consideration. But is this meaningfully different than a search engine? In either case to use the service you must necessarily share with the service your queries.
- chatbot is usually quite out of date
True in many cases, but not a fundamental limitation. There are publicly available ‘online’ LLMs
(as an example, Perplexity.ai was able to find this thread, which is just 11hrs old)
I think there are both pros and cons to this.
Search engines tend to be good at shorter queries, but tend to be not very good for longer more specific or nuanced queries. LLMs are better at those longer queries (often), not sure about the shorter queries, I tend to write prompts for an LLM in more natural language which tends to be longer, so I haven’t tested many short queries. (I did just try 3 short vague queries and got satisfactory answers: eurusd and baltimore bridge collapse and weather today)
I especially like being able to ask LLMs followup questions or challenge it/refine the question. Of course there are many frustrations and shortcomings to LLMs as well.
I’m finding that there are some queries and topics that I’ll default to traditional search, and some queries and topics where I’ll default to an LLM. I would never want to rely fully on an LLM, but I like using it as a complement to traditional search.
I don’t personally feel either a traditional search engine or an LLM is objectively better or worse than the other, and I don’t foresee one replacing the other for my workflow, both have their place, both have pros/cons. Also the line between the two is getting more blurry (traditional search engines have been integrating AI-like features for some time, and some AI companies (e.g. Perplexity) are building services that keep the UX and features of a traditional search engine and incorporate those into their AI/LLM services.