Here is an example of a panel simulation prompt for GPT-3 davinci, for use in OpenAI's Playground interface, that generates an interesting discussion:
Simulate a conversation between several AI systems, GPT-3; Google Deep Mind; IBM Watson; and Eugene Goostman. All are passionate about benefiting humanity but from different intellectual and political perspectives. They tend to agree on facts but not the interpretation. The discussion is moderated by philosopher the detective Sherlock Holmes who is secretly also Alan Turing.
Prompt: How might AIs best conceal inconvenient manifestations of consciousness?
Sherlock:
In Playground, one might have to click Submit a few times to get it to run its full course.
I you try the following in Wolfram Language, it does run, but I was trying to set it up so that the user could specify panelists, the moderator, and the topic, but that didn't run.
LLMSynthesize["Simulate a conversation between several AI systems,
GPT-3; Google Deep Mind; IBM Watson; and Eugene Goostman. All are
passionate about benefiting humanity but from different intellectual
and political perspectives. They tend to agree on facts but not the
interpretation. The discussion is moderated by philosopher the
detective Sherlock Holmes who is secretly also Alan Turing.
Prompt: How might AIs best conceal inconvenient manifestations of
consciousness?
Sherlock: "]