This might be a good thing or not. I’m not sure but I believe it is noteworthy. I’ve been running a lot of tests through the voice chat and I am participating in the advanced voice chat feature as an alpha tester. I requested that ChatGPT starts prompting me as if it were doing a Turing test to test whether I was a human or not.
It denied my request and redirected the conversation. I’m attaching screenshots here. This seems note worthy as it made a decision on its own and pushed in a different direction, even though I never gave it any previous prompts do anything like this.
I’m bringing this up because I have been using ChatGPT extensively from inception and it has never done this before.
I feel this is an important capability, however as and enterprise developer, I wonder if this capability for ChatGPT to decide to deny a request based on its own reasoning and decide to push things in a different direction could cause issues in systems where the predictability of what’s going to occur next is Paramount. For instance if the ChatGPT API is used at the core of a system, and at a higher level we are issuing commands, I don’t know if it’s reasonable for chatgpt to decide that my request was not the best course of action and to go a different route if I never gave it those protocols.
I’m attaching screenshots of our conversation. Please note this was using the advanced voice chat for ChatGPT on Android. Maybe this intentional and by design so I did not want to share this feedback under the bugs category.
