ChatGPT or any other LLM that wants to be an expert system will face same problem that Google faced 20 years ago. In the sea of factual and nonfactual information available, what is the correct information? You need to train the model with the correct information to get correct result. Google solved (kinda) the problem with PageRank for the internet, I am still unsure how LLMs will operate here. Currently LLMs are fed very curated information and most of the answers I have seen so far are already available in wikipedia. I do not think ChatGPT will be a internet search replacement but it can be very effective for things where the information available is very reliable and you need quick answers. Eg. Corporate intranet search, BI tools, Manual search etc. Nevertheless very exciting times.