AI hallucinations: ChatGPT created a fake child murderer | Hacker News Reader

AI hallucinations: ChatGPT created a fake child murderer

noyb.eu

14 pointscroes9 comments

Maybe there needs to be a disclaimer on all Gen AI outputs:

This is a work of fiction. Names, characters, businesses, places, events, and incidents are either the product of the model's imagination or are used fictitiously. Any resemblance to actual persons, living or dead, or actual events is purely coincidental.

Stop asking generative AI for "Information" and ask it to generate ideas. I guess there needs to be a basic disclaimer and and 1 line tutorial before every query.

This is a stupid lawsuit and I hope noyb loses it.

If I make a website where you press a button and get a random string, should I worry about being sued under gdpr for defamation? Particularly when I make it clear it's getting a random string?

I should hope not, but somehow noyb thinks otherwise.

It’s not random strings and people believe Chatbots.

Maybe you think differently if it’s you and the people who ask the Chatbot about you are your neighbors or employee.

I have asked it about myself, it got some stuff right and most of it wrong, just as expected.

It's little different from rolling a dice to pick the next word, should that be made illegal? Of course not.

The way these llms works are extensively documented and published, there are thousands of scientific papers about it, and even if you are not inclined to that you can find hundreds of youtube videos explaining how it works.

Their very disclaimer states it should not be relied upon as a source of truth.

Anyone who cares to know, knows that.

If this guy didn't, it's because he didn't care to know.

Except that OpenAI is very much trying to establish their products as being worthy of credence, disclaimers notwithstanding.

Both the disclaimers and the scientific literature say the same thing.

Just because you want to ignore it doesn't mean it doesn't exist

My point is a little different. OpenAI is pushing its ChatGPT product as a solution to real-world problems that the company knows it cannot solve reliably. There are some minimal disclaimers, but they do not dominate the messaging: the marketing does. OpenAI is fundamentally interested in selling the product, not promoting responsible use of the product. So the mixed messaging is calibrated so that lay users are encouraged to rely on ChatGPT even when they shouldn't. Kind of like Full Self Driving, which isn't.

Hallucinated murderers shall be convicted under hallucinated laws. Mission accomplished?

Reply on news.ycombinator.com