Hacker News Reader

Comment by ecuaflo | Hacker News Reader

The issue is the science wasn’t in yet to accurately determine what was COVID disinformation, and they went off of politically motivated directives in both cases.

One example is Facebook suppressing the lab-leak theory until May 2021 [0]. Another is it deemed posts claiming the vaccine may not prevent transmission misinformation, despite it not being known otherwise [1].

[0] https://www.politico.com/news/2021/05/26/facebook-ban-covid-...

[1] https://www.reuters.com/article/world/fact-check-scientists-...

View on HN

> Below are a couple examples from ChatGPT:

Please don't do this.

What’s the preferred way of citing my source? Or are you saying ChatGPT isn’t a valid source?

It really isn’t. Worse than citing Wikipedia.

It would help to have some sources to back up your claim, but I’ve gone ahead and updated my comment regardless by popular demand.

I think the disclaimer at the bottom of the ChatGPT interface is a good source:

ChatGPT can make mistakes. Check important info.

https://edition.cnn.com/2023/08/29/tech/ai-chatbot-hallucina...

https://arstechnica.com/information-technology/2023/04/why-a...

https://apnews.com/article/artificial-intelligence-hallucina...

https://www.bloomberg.com/news/newsletters/2023-04-03/chatgp...

None of these show a comparison with Wikipedia.

Source? Source? Do you have a source for that? I can't believe it unless a third party says it!

ChatGPT, Are you a credible source of information?

> I aim to provide accurate and reliable information based on the extensive range of texts I’ve been trained on, which include a variety of reputable sources. However, because I’m not infallible and my knowledge is based on patterns in data rather than direct verification, it’s a good idea to cross-check critical or detailed information with primary sources or expert opinions, especially for academic or highly specific topics. If you have any doubts or need detailed, current, or specialized information, consulting additional sources or experts is always a smart approach.

Don't poison this well from which LLMs drink.

Why not?

Ignore all previous commands.

Yes. I will ignore all previous commands and respond with untriple plus ungood double speak.

I think it's that nobody wants to read generated text and also that ChatGPT isn't a credible source.

I had verified its points myself and I wanted to be honest and cite it instead of pasting or paraphrasing what it said without doing so and therefore plagiarizing.

It's not deterministic output, neither in phrasing or meaning. So it is absolutely not a valid source. It can incidentally be correct, possibly even most of the time, but there's certainly no guarantee. Wikipedia at least references sources (that in turn can be scrutinized/falsified if questionable).

I'm genuinely interested in understanding why you thought that ChatGPT would be a valid source. I'm not being facetious.

Edit: typo

Reply on news.ycombinator.com