Twitter user reverses roles with ChatGPT

Updated on: 26 June 2023

Justinas Vainilavičius
Senior Journalist

A developer has reversed roles with ChatGPT to answer the bot's "dumb questions" instead of the other way around.

Developer and designer Nate Parrott has shared his "little chatbot experiment" on Twitter, where he posted a short video clip of an exchange.

The conversation starts with Nate typing in "hi" in the chat window and the bot responding with the typical "How can I assist you today?"

Usually, this is the part where the user asks a question, but Nate presses the button saying "reverse roles," and now he's the one under the chatbot's interrogation.

"What's your favorite book," ChatGPT demands, to which Nate types back – taking time to correct the typos – that he is a large language model (LLM) and, as such, doesn't have one.

"But people like Harry Potter," he adds and is confronted with the follow-up question barely a fraction of a second later – what is Harry Potter?

Slowly but surely, Nate explains it is a story about a boy who learns he is a wizard and goes to a wizarding school.

twitter

That is not enough, and the bot wants to know more, which is too much for Nate as he presses the reverse button to have ChatGPT answer its own question.

After that, Nate reverses roles again and claims credit for the synopsis. ChatGPT does not appear to mind but has no further questions.

Reactions to the experiment varied from "lol" to "got me in a thousand-yard stare wondering if I'm anything more but a[n] LLM."

"Humans are back, baby!" one comment read, to which Nate responded, "ngl [not gonna lie] this is hard work."