Agree. I tried the first 3 examples:

* "Rubber bouncy at Heathrow removal" on Google had 3 links, including the one about SFO from which chatGPT took a tangent. While ChatGPT provided evidence for the latest removal date being of 2024, none was provided for the lower bound. I saw no date online either. Was this a hallucination?

* A reverse image lookup of the building gave me the blog entry, but also an Alamy picture of the Blade (admittedly this result can have been biased by the fact the author already identified the building as the blade)

* The starbucks pop Google search led me to https://starbuckmenu.uk/starbucks-cake-pop-prices/. I will add that the author bitching to ChatGPT about ChatGPT hidden prompts in the transcript is hilarious.

I get why people prefer ChatGPT. It will do all the boring work of curating the internet for you, to privde you with a single answer. It will also hallucinate every now and then but that seems to be a price people are willing to pay and ignore, just like the added cost compared to a single Google search. Now I am not sure how this will evolve.

Back in the days, people would tell you to be weary of the Internet and that Wikipedia thing, and that you could get all the info you need from a much more reliable source at the library anyways, for a fraction of the cost. I guess that if LLMs continue to evolve, we will face the same paradigm shift.