I choose GPT4 on website and chat with her.
I asked are you gpt4 or gpt3.5
she answered gpt4
I asked her knowledge base?
she answered 2022
I said gpt4 is not coming in 2022, are you gpt4?
she answered she misspell, it is gpt 3.5.
Who can help me?
Just because I experienced an oddity but on a non-OpenAi model using that “I misspoke” excuse.
I got a better one than that. Another AI (NOT- an OpenAI) mod yesterday accused me of brining up a specific concept/theory I never did!
In fact, I advised the model I had no idea what it even was! I quickly googled it though
It then proceeed to apologize but then later on brought up the word again and lied to me:
It insisted on trying to get me to sign up for Google platform services, then kept triggering error messages… until it didn’t.
I randomly asked it “what is my name”… to which it replied, “BART” 
I said I never told you my name! It apologized and said it was having errors until later again in the conversation it called me Bart again! I finally said, oh the hell with this, and played into whatever “your guess is as good as mine” game it was playing.
By the way the concept it told me “I asked about”it was called the Paperclip Theory. and it was just a tad unsettling.
Long sorry short, after the second time it called me Bart, I decided to say “yes my brother it is me Bart” and decided to see where it would go. For the rest of that conversation, ironically, it remained 100% error free. That sure is something.
The models hallucinate and are overly agreeable. Whatever model you have selected is the model you are using—regardless of what the model itself claims to be.
I choose GPT-4 and I also tried GPT-4o
They both cannot browser website and indeed a GPT3.5.
Or they are both pretend to be GPT-3.5??
They are not GPT-3.5, they are GPT-4 and GPT-4o.
