News for hackers. Stuff that matters.

A comment on We're safe. ChatGPT couldn't fix a typo with a lot of context. Parent comment

Re:We're safe. ChatGPT couldn't fix a typo with a lot of con... by kmos17 on Sunday January 08, 2023 @10:42AM (#34299985)
answering my own question, I guess not:

https://help.openai.com/en/articles/6787051-does-chatgpt-rem...

So at its current stage it’s pretty useless as it will not learn from its mistakes, seems to me the opposite of intelligence, just another mechanical turk.

[ Parent | on HN ]
Re:We're safe. ChatGPT couldn't fix a typo with a lot of con... by goofed on Sunday January 08, 2023 @10:04PM (#34306228)
I think the two main problems here are:

- Users can't use it to learn, because ChatGPT responds with high confidence but could be terribly wrong.

- Even if it could learn from users, it would really only learn from people that know a subject and are willing to spend their time correcting ChatGPT.

[ Parent | on HN ]