I Interviewed ChatGPT About AI Ethics — And It Lied To Me

Abstract

ChatGPT is remarkable. It’s a new AI model from OpenAI that’s designed to chat in a conversational manner. It’s also a liar. Stuck for ideas on what to talk to a machine about, I decided to interview ChatGPT about the ethics of AI. Would it have the level of self-awareness to be honest about its own dangers? Would it even be willing to answer questions on how it behaves? Yes, it would. And while ChatGPT started off by being commendably upfront about the ethics of what it does, it eventually descended into telling outright lies. It even issued a non-apology for doing so. Here, then, is the transcript of my interview with ChatGPT. I’ve omitted some whole questions, as they’re frankly too boring to include, but every answer here is the full answer provided by ChatGPT. Stick with it until the end, when things take a very strange turn...