News for hackers. Stuff that matters.

A comment on Teaching GPT-3 to Identify Nonsense. Parent comment

Re:Teaching GPT-3 to Identify Nonsense by blixt on Wednesday July 29, 2020 @04:26PM (#23991188)
These examples I gave were on a pretty high temperature (which I think makes the fact that its answers are correct even more interesting), so YMMV. Here's a gist with the full transcript (note that several of the items at the top are written by hand, not output by GPT-3):

https://gist.github.com/blixt/b48d2d41590d9b4ad2faeb66d63c3f...

I'm also sure there's a better set of input Q/A to make sure it's able to cover more question formats.

[ Parent | on HN ]
Re:Teaching GPT-3 to Identify Nonsense by longtom on Wednesday July 29, 2020 @04:33PM (#23991281)
Which parts were written by a human?
[ Parent | on HN ]
Re:Teaching GPT-3 to Identify Nonsense by skybrian on Thursday July 30, 2020 @05:21PM (#24002969)
The first question I got good results for was:

  Q: Who did Capablanca defeat to become world chess champion?
I used everything before that as a prompt and didn't have any luck with a shorter prompt.
[ Parent | on HN ]