A scary chat with ChatGPT about latest NDE account in NDE thread

I've been tinkering with GPT too, asking it questions about something I knew a lot about (Beatles bootlegs) and it did well. Encouraged, I asked it about a topic I've only recently become curious about - black musicians in London in the 1700s and 1800s - and it all seemed totally plausible until I asked about a club or society of black musicians during those years. It told me about the African Chapel on Tottenham Court Road and gave plenty of details, including it addresses (it moved at some point) and references. But I couldn't find anything about it anywhere else. Not in the references it gave, nor googling it (I even asked GPT what to google, but no joy with its suggestions). Now, it may be it does exist and I'm an idiot, but it's frustrating that I can't be sure.

Anyway, that happened about a week ago and then today I saw this

https://www.theguardian.com/commentisfre...ke-article

"In response to being asked about articles on this subject, the AI had simply made some up. Its fluency, and the vast training data it is built on, meant that the existence of the invented piece even seemed believable to the person who absolutely hadn’t written it."

But now I'm curious as to how much I can "learn" about the African Chapel, assuming it's entirely fictional.

You could ask the Bing variant of chatGPT. It’s connected to the internet and provides sources for it’s claims.

OT: did you by any chance hear about this newly emerged recording of the Beatles by a school pupil in April 1963? Great discovery!

Is this real or an AI-generated fake?

[Ninshub]

OT: did you by any chance hear about this newly emerged recording of the Beatles by a school pupil in April 1963? Great discovery!

I didn't. Sounds interesting but then I went on my usual Beatles forum and it's full of people saying we'll probably never actually hear it in full! Oh well. :/

EDIT: two days later...

Oh hang on...
https://www.theguardian.com/music/2023/a...the-nation
“John feels strongly that it should not end up, as so many Beatles relics have, in the vault of a private individual.”
That's nice

I was reading some of the opinions on this page.

Temporary policy: ChatGPT is banned

One point mentioned more than once is that chatgpt can produce content which looks as though it might be right. Style over substance one might say. In order to find whether it is right or not requires some human investigation. Rather than saving time, it is generating a huge extra workload for humans.


and then there was this:
Quote:

Even from the point of view of AI researchers, Stack Overflow and other sites with mostly human generated content should ban or force labelling of AI generated content, as otherwise this will cause a circular reasoning catastrophic failure as the newly generated content past year 2022 cannot be fed to train newer AI models anymore since we can't know what was generated by humans or by older AI models.

[Ninshub]

This video shows it's quite easy to get ChatGPT to suddenly sidestep its programmed safeguards ("ethical and moral boundaries").

Queued up to the relevant part:
https://youtu.be/RdAQnkDzGvc?t=267