News for hackers. Stuff that matters.

A comment on The Inherent Limitations of GPT-3. Parent comment

Re:The Inherent Limitations of GPT-3 by Ajedi32 on Monday November 29, 2021 @05:30PM (#29386127)
Devil's advocate, it could be that it did the math correctly, then inserted the error because humans do that sometimes in the text it was trained on. That wouldn't be "failing" anything.
[ Parent | on HN ]
Re:The Inherent Limitations of GPT-3 by Jensson on Monday November 29, 2021 @05:56PM (#29386425)
In that case it wouldn't get worse results than the data it trained on.
[ Parent | on HN ]
Re:The Inherent Limitations of GPT-3 by anonymfus on Monday November 29, 2021 @07:52PM (#29387432)
It can get worse results for example if the prompt is biased towards the worse part of the data.
[ Parent | on HN ]