Hi everyone,
In the GPT:
-
the model gathers the call parameter from the conversation and passes it to the action’s API call. So, could the model hallucinate in the passed data? If I asked GPT to book a ticket to Paris, could it book a ticket to London?
-
The response of the action’s API call is delivered to the model then the model passes it to the conversation. So could the model hallucinate in the response? If the action brings SEO information from Ahrefs, could the GPT fake some information?
Are there resources that include discussion or affirmations on the topic?