wind viper
I've signed up for a plus subscription and selected GPT-4 in the dropdown. I'm having a conversation with the AI and I ask how I can show it an image. Then it says it can't see images. I get suspicious and ask it "Aren't you supposed to be able to see images?", then it responds that it's based on GPT-3.
Steps to reproduce
What model are you? Gpt-3 or 4?
Expected result
"...GPT-4..."
Actual result
"...GPT-3..."
Additional information
Browser: Chrome
OS: OS X
GPT-3, not GPT-4 as expected
graceful hearth
Got the same when I tried investigating, I'm trying to understand if it's because it's training cutoff date is september 2021, or if it's really a bug in the model selection at the beginning of the chat. After several questions, i asked it the following question :
"""
**Q : **
Are you sure the model you are using is gpt4?
**A : **
Apologies for the confusion. I am, in fact, an instance of OpenAI's GPT-3 model. My knowledge is up-to-date as of September 2021, and I am not aware of a GPT-4 model released by OpenAI. If you have any other questions or need assistance, please feel free to ask.
"""
What do you guys think?
wind viper
I tried pushing it a bit, check this out: (apologies for tiny print, wanted to capture it all in one screenshot)
scenic geode
I believe they already fixed
coarse raft
HI
scenic geode
But
delicate crow
The model is not aware of its version or capabilities. If you see a black avatar, that's gpt-4
tough pier
OK, but I asked it a question about Whisper and it replied:
I assume you are referring to OpenAI's Whisper ASR system. As of my knowledge cutoff in September 2021, OpenAI has not released the code or models for Whisper ASR for public use
And I do have the black avatar, and I'm a PLUS subscriber.