Hmm. Well, I find chatGPT similarly useful to Google even in this early stage, albeit for very different types of question. I'd say it's more likely to disrupt WolframAlpha than Google, if my usage is typical — while chatGPT does make basic errors 5% of the time, WolframAlpha just straight up doesn't understand my query 33% of the time.

ChatGPT and other LLMs will probably also disrupt recipe blog spam, with potentially hilarious results for all foods and drinks whose names are double entendres.

LLMs will also, if I correctly understand the difference between opinion columns and journalism, the writing of opinion columns in newspapers. And turning dry research into colourful column inches.