One thing I hate about chatgpt

[Isaac]

....is when it flat-out lies to me. coming back saying "Ok, now I've expanded this document and it is over 3000 words". I open it in word, and it's 1100.

I know it's capable of counting words, because if I ask it to count words it will do it correctly.

It just makes stuff up some times - so frustrating!

 

[jdraw]

It just makes stuff up some times - so frustrating!

A few other observations:
Changes topic/context direction;
Hallucinates (makes up functions/terms that do not exist);
Offers enhancements before completing current situation;
Changes response format even after being requested to offer replacements/rather than piecemeal fragments;
Often forgetful during session .......

 

[GPGeorge]

All of that stuff! Often, when trying to trouble-shoot a specific problem, Claude or Chatty start offering additional components beyond the one I'm working on. And at some point, I get lost in a series of responses, some of which are on topic, some of which are looking down the road. It gets confusing about what is relevant at any given point. I just want to resolve a specific issue and then move to the next one, not try to guess which of three or four previous responses remain to be considered.