GPT-4o answers incorrectly about its own version

Question

GPT-4o answers incorrectly about its own version

GPT-4o answers incorrectly about its own version. Are you experiencing the same issue? This chat interacts with the GPT-4o model's API from OpenAI. This can confuse users who might believe it's not the real GPT-4o. Additionally, I've noticed inconsistencies in its answers at times.
https://asset.1min.ai/images/Screenshot%202024-06-07%20at%2010.48.21.png

ksaj · Accepted Answer

Thinking how the answers are derived, are you surprised? People seem blind to the word "artificial" and try to give it properties that just aren't there.

I gave exactly the same prompt and got:

  > Yes, I am based on the GPT-4 architecture, often referred to as ChatGPT-4. How can I assist you today?

So it can answer correctly sometimes (probably often).

I asked Google the same question. It's not AI, but we have result accuracy expectations with it that match yours for ChatGPT. It doesn't seem to be aware of versioning at all. So, I wonder how important this is to a user. Why would a user care, let alone doubt, as you asked? Afaict nobody has ever pestered Google for not putting version numbers on their results.

Users want results. Not version numbers and technical stuff that belongs underneath and with the developers. Understanding a little bit about how results are derived in ChatGPT means this sort of error should be expected, and not considered a singular "issue."