smoky zenith
Steps to reproduce:
- Usage of GPT 4model only. While there isn't a clear pattern to report, (and tried to make it happen), they are causing disruption in the functionality and usability from 40,50,60% of daily chats. ...that much.
Expected result:
Expected: Should adhere to both custom setting and MyChat builder setting and manual inputs. For tone and specifics sets such as no closing comments etc. and requests on tasks.
Actual result:
Consistently disregards user instructions, including failing to omit closing statements and altering provided text despite user requests even after multiple corrections in the same conversations.
Additional information
Since the introduction of MyChatGPT: Frequent retraining model even on same conversation, similar to experiences I encountered back in Jan 2023. Acts like GPT3??Deleting and starting a new conversation temporarily removes bug for a few outputs. Refusals, rejection, avoidances are on and off, same for web searches, with message -no capable to search and prematurely cutting off just after a few "fresh" interactions.
NOTE. Does not happen with messages that were created before the intro of myChat.Browser: Safari, Chrome,
OS: iPad Pro 17.2, Mac M2, IMac,
unkempt wraith