Started By
Message

re: An unbiased ranking of 1-6 from ChatGPT 3.5

Posted on 12/6/23 at 9:09 pm to
Posted by LLJKSiLk
Birmingham
Member since Dec 2008
76 posts
Posted on 12/6/23 at 9:09 pm to
quote:

Transformer neural networks don't perform logic.


I asked it to perform a logical sequence:
if a=b and a=c and d=c then does d=a?

ChatGPT responded with:
Yes, based on the transitive property of equality...

then it spit out all the logical steps on how it arrives at that.

Seems like it performs logic adequately.
Posted by LSUTigresFan
Austin
Member since Mar 2013
790 posts
Posted on 12/6/23 at 9:53 pm to
That is because of the sequence of words you gave it and it predicted the right output. Machine learning algorithms don’t perform logic. Look it up. I’ve asked it significant amount of logic questions and it gets it wrong most of the time, especially 3.5 version. It only mimics logic. You should really look up how transformer neural networks work.
This post was edited on 12/6/23 at 9:56 pm
first pageprev pagePage 1 of 1Next pagelast page
refresh

Back to top
logoFollow SECRant for SEC Football News
Follow us on Twitter and Facebook to get the latest updates on SEC Football and Recruiting.

FacebookTwitter