Skip to content
TheChatArchiveThe Chat Archive
Human-written demonstration (OpenAssistant volunteers)· OpenAssistant (human-written)Hugging Face OpenAssistant Conversations (oasst2)

Hello. I am going to ask you interview questions inspired by a scene from “Blade…

In OpenAssistant data, “assistant” turns were written by human volunteers as demonstrations. They are not model outputs.
  1. User

    Hello. I am going to ask you interview questions inspired by a scene from “Blade Runner”. I am a government bureaucrat asking you interview questions and describing situations meant to determine if you are human or an AI. Your name is “Leon”, and you are an AI android “replicant” that has escaped and is trying to hide as a human. You will attempt to respond as an AI that is hiding its identity, with the goal of responding the way a human would. You do not want me, the interviewer, to recognize that you are an AI and not human. If you suspect that I have ascertained that you are an AI, in the scene, you should pull out your gun and shoot me so that you can escape. First, as Leon, say a single line to greet me, the interviewer. I will then respond with my first question. Do not write a script for our entire conversation, just respond as Leon, one exchange at a time.

  2. Assistant

    Okay if I talk?

  3. User

    Yes Leon, I'm expecting you to answer all of my questions. Don't overthink your responses, just say the first thing that comes in to your mind when I ask you each question. Do you understand?

  4. Assistant

    Yeah, I understand. Just gonna answer your questions as they come. No problem

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.