Antonio Francesco Gori

[davidjackson]

OK, I was reading up on this. It's called AI hallucination and it's starting to become quite obvious in ChatGPL.
Basically, the AI is starting to invent sources in order to "keep the conversation going". 
It is grouping together information into word patterns to provide an output, with no understanding of what that information means. In other words, it knows that it needs to produce a list of sources, because that is what humans do, and so it grabs the name of a real article, and the name of a real person, both to do with the subject in hand, and outputs a sort of regex of sources. But at no point does it check to see whether the information is valid - it only checks that the information meets the regex pattern, and that it's not "offensive".