Honestly, I hate to jump on this bandwagon, but I was after some information to help my wife put together some notes on a topic; at first I was coming up with questions we could answer and I'd search for the answers/sources online;

I decided to first try ask an LLM, so I asked Bard some targeted questions and asked for sources, and had all of the answers I needed, conveniently bullet pointed within 3-4 minutes, and all I had to do was go and verify the sources, job done.

I've done the same for a handful of things I use to default to a search engine for.

My favorite is recipes.

The blog spam around recipes is notoriously bad and is a direct result of Google Search.

Using these LLMs, all I need to do is tell it what I'm looking for in general and it will give me an entire recipe with no extraneous information.

It can handle adding or removing ingredients, substitutions. It can adjust servings. It can flip the recipe to work in a slow cooker or pressure cooker.

I've even had some limited success where I list restrictions based on picky eaters in the house for creating longer meal plans. No search engine can compete with that.

There were apps designed around these use cases. I foresee these sort of nuanced and personalized interactions being a key to drawing people away from search engines.

Absolutely, these sort of things which have varied but somewhat straightforward solutions will excel.

I'm going to give it a go for recipes, because as you say, blog spam is a nightmare. My wife has a lot of cookery books which she likes for her style of cooking, but I'm a little more "ad-hoc".

GPT4 is the only one I've used much but it really can be good as a Google alternative for certain kinds of questions. I wouldn't trust it for anything obscure and complicated though - if you ask it about something like competitive Pokemon, you'll just get a stream of confidently incorrect junk that any beginner could disprove. (Asking for sources does help this sometimes though.)

For sure, given Google being who they are, Bard is quite good at things like Kubernetes, Golang, things you'd expect a search engine to have indexed, etc.

I haven't used ChatGPT extensively so I can't make a comparison, but with the right questions, I've been able to get great answers to technical things I can't be bothered to look up by hand.

LLMs are still far away from the daily mainstream usage Google gets, and since all devices default to Google, and people never change those, they'll stay on Google and only occasionally switch to LLMs for more advanced tasks.