Following on from my previous poll, I asked ChatGPT about MACE vs MACB timestamps.

It has basically invented the answer. The work it cites is a real document, written by Dan Farmer and Wietse Venemea, but it was a Usenet post in 1993 and made absolutely no references to timestamps.

https://www.dcs.ed.ac.uk/home/rah/Resources/Security/admin_guide_to_cracking.pdf

It seems like ChatGPT made up an answer and then found an approximate reference to make it look legitimate, without checking what the reference included.

To this extent, ChatGPT is basically like most humans in a discussion...

@tazwake

This is not how chatgpt works. It's not checking references. It's a fancy text generator. Thus, it does not check references nor does it check facts.

@13reak I totally get that it is really good at random text generation. As I said, it is like most humans in a discussion.

Its a shame so many people think it does more though.