Impromptu: Amplifying Our Humanity Through AI This metaphor breaks down when it comes to tools like GPT-4. Rather than manipulating a physical tool to generate a predict- able and planned result, using GPT-4 is like directing a scene from a movie. As the director, you’re working with an actor to elicit the best performance. You’re not telling them, “Bend your neck fifteen degrees, and then after 2.5 seconds, look at the person across from you.” Instead, you’re asking them to make the audience feel a certain way: “Convince us you’re in love.” GPT-4 isn’t a human actor, but the process of directing it is similar. You have to coax out your desired result, and you may need to try many variations on a prompt in the same way a director might need to film multiple takes of the same scene. It’s ultimately a collaborative process. Principle 3: Just try it! In most of our work, we’re taught to plan in advance and avoid making mistakes. That’s because implementing a plan is costly in terms of time and other resources—there’s a reason a car- penter’s adage is to measure twice and cut once. But what if implementing a plan was cheaper and faster than thinking about it? That’s the confounding paradox of GPT-4 and LLMs. In far less time than it takes to debate a plan, GPT-4 can simply gen- erate a complete response for you to review. If you don’t like the response, you can throw it away and generate another one, or you might just generate three variations to give you more potential choices. 134