ChatGPT AI lies

[satyr85]

Lawyer cites fake cases generated by ChatGPT in legal brief

https://www.legaldive.com/news/chatgpt-fake-legal-cases-generative-ai-hallucinations/651557/

 

[awesomewebsites]

Whose fault is this? Not the chatgpt for sure.

 

[roydan]

100% human written depositions at $1/1000 words

 

[D Ran]

This is pretty old news. Open AI states not to use GPT-3.5/4 generated stuff for serious uses if you do, double-check the work. He took made-up case law that ChatGTP conjured up to argue in court (note the case law sounded convincing). The judge was like I have never heard of such cases, and he had to fess up that he used ChatGPT. It's a good eye-opener for many who rely solely on ChatGPT for serious work. Dudes gonna get disbarred for this; it's kinda hilarious. GPT hallucinates a little too much still, even GPT-4. For context, there is a 10% hallucination rate on GPT-4.

 

[medialounge]

Lawyer cites fake cases generated by ChatGPT in legal brief

https://www.legaldive.com/news/chatgpt-fake-legal-cases-generative-ai-hallucinations/651557/

In the future, it's possible that ChatGPT could play a role in legal proceedings, reducing the need for a lawyer as a representative might be sufficient to present a case.

 

[imanseptian]

I wonder, in the future, a lazy doctor consult chatGPT to treat the patients.

 

[Gogol]

I once gave it a prompt to generate some code for me using a well known npm plugin. It came up with completely imaginary method that doesn’t even exist in that plugin. When I pointed that out, it was like, i am sorry, you’re right. Then it came up with another imaginary method lol.

Best of luck having to depend on chatgpt for anything hahaha.

 

[Spicy Chromosomes]

I wonder, in the future, a lazy doctor consult chatGPT to treat the patients.

in the future? i'm sure that there are some doctors doing that right now

 

[TZ2011]

Lawyer cites fake cases generated by ChatGPT in legal brief

https://www.legaldive.com/news/chatgpt-fake-legal-cases-generative-ai-hallucinations/651557/

Are you using Internet Explorer? Happy new 2021

 

[Gitton]

This is pretty old news. Open AI states not to use GPT-3.5/4 generated stuff for serious uses if you do, double-check the work. He took made-up case law that ChatGTP conjured up to argue in court (note the case law sounded convincing). The judge was like I have never heard of such cases, and he had to fess up that he used ChatGPT. It's a good eye-opener for many who rely solely on ChatGPT for serious work. Dudes gonna get disbarred for this; it's kinda hilarious. GPT hallucinates a little too much still, even GPT-4. For context, there is a 10% hallucination rate on GPT-4.

chat gpt 4 is better? I can't find it

 

[MoonStarz]

a couple of days I read that someone was going to sue the devs of chatgpt because the AI said he was under investigation by the FBI because of white collar crimes but the funny thing that wasn't true, but that make me thing, if that information is feed by chat gpt manually from where did they get it in first placed? is there any leak document or was a news that was never released to public like a conspiration!!!! the AI can't create fake new so that story is some where in the internet and maybe is the same on this case.

 

[PURSUIT MEDIA]

The way users are misusing ChatGPT each day is alarming. Now who is at fault?

 

[FBUnlim]

Such incidents are unavoidable. Problems arise, problems get fixed—I believe this won't be an exception.

 

[Crucix]

I work in tax and asked tax based questions relevant for Australians. It gave me answers to tax based on a US citizen (completely wrong) that won't apply in Australia.
AI is only good for general questions and even then you have to double check (atleast for now)
I wouldn't rely on it for anything technical

 

[Alditio]

At one time chatgpt gave me a wrong birthday for a president and I had to check with it twice to get the real answer and when confronted as to why it gave me a wrong answer the first time all it can do is just apologize

 

[pj777777]

And they say gpt isn't good for writing fiction lol

 

[smudge_dove]

is there any leak document or was a news that was never released to public like a conspiration!!!! the AI can't create fake new so that story is some where in the internet and maybe is the same on this case.

There's actually people out there who take anything an AI text model produces at face value
Worse, even: there's people out there who take anything a human says at face value lol

Makes ChatGPT a lil human, doesnt it

 

[kingfer]

GPT3.5 The things he can do are very limited. If you ask him to do math problems and ask multiple times, he can provide different answers

 

[doubledhamal]

In the future, it's possible that ChatGPT could play a role in legal proceedings, reducing the need for a lawyer as a representative might be sufficient to present a case.

No that will not happen. We cannot play with the lives of people with auto generated text.