amberage, 1 month ago

@noybeu

On the contrary, generative AI tools are known to regularly “hallucinate”, meaning they simply make up answers.

That's all they do. Even when they show correct information, models have no concept of what that means, what a person is. They calculated the most likely sequence of words, and it just happens to be correct sometimes.

revk, 1 month ago

@noybeu I would imagine the problem is that is does not claim it is accurate information, or that it is about the actual person you ask it. It is fiction, and if it claims to be fiction, it can't be "inaccurate personal information" or even "person information", can it. It I say Granny Weatherwax is dead I am not claiming to be stating accurate information of someone using that name, and can't be expected to "correct" it.
Surly.

waggers5, 1 month ago

@noybeu It's a language model. It's not a source of data.

It's like trying to sue the English language because it's possible to use it to say something defamatory.

ChatGPT's ONLY job is to simulate the structure of, e.g., a conversation. Managing the content of that conversation is beyond its scope.

Substitute "conversation" for letter, essay, lecture, etc.

WhyNotZoidberg, 1 month ago

@waggers5 @noybeu hence you sue the developers of said LLM. Because just like "English" someone is telling it how to behave and respond.

The developers of any given AI are responsible to what it does, just like they are responsible for say copyright violations while training it.

wizzwizz4, 1 month ago

@waggers5 @noybeu Except (a) that's not the product that OpenAI is marketing; and (b) even if it were, the model does contain personal information about many people. (GPT-3.5 Turbo answered "What's Beyoncé's birthday?" with "Beyoncé's birthday is on September 4th." (And now I can't say "I've never used ChatGPT" in internet arguments: I hope you appreciate this sacrifice.))

allo, 1 month ago

@wizzwizz4
Aren't they displaying the disclaimer that one should not rely on the output anymore?

I did not use it for some time (local models are getting real good and preserve your privacy) but the last time I used ChatGPT they basically told me not to believe anything it says without checking and not to input any privacy related data.