Discover more from One Useful Thing
How to... use ChatGPT to boost your writing
The key to using generative AI successfully is prompt-crafting
I think most people who are using ChatGPT to help with writing are doing it wrong. I don’t just mean because they using it to cheat on school assignments (don’t do that) or because they don’t check the facts that ChatGPT gives (they might be made up), but because they have the wrong mental model for how to work with the system.
I have mentioned that ChatGPT isn’t Google, and it isn’t Alexa, but it also isn’t a human that you are giving instructions to. It is a machine you are programming with words.
You are writing a prompt, not having a conversation
Because ChatGPT often acts like a helpful human, we get lured into thinking it is one. I asked my Twitter followers, and nearly half always act politely towards the AI, and another quarter mostly act politely. I do this myself, saying “please” and “thank you” in my requests, even though the AI doesn’t care.
While there is nothing wrong with politeness, it often obscures the fact that all we are trying to do is prompt a non-sentient machine to generate the text we need. This confusion isn’t a problem in other types of generative AI. Since AI image generators are not as verbose as ChatGPT, people using them have learned they need to spend time doing “prompt engineering:” editing and playing with prompts to come up with the results they want. These prompts can get very elaborate, and can often seem abstract, or almost poetic. They are the program the AI is following, where the goal is to generate a great image through trial-and-error and pushing the AI in the direction you want.
But between the human-like nature of chat and the fact that written material is harder to immediately evaluate, many people tend to avoid explicit prompt construction in ChatGPT. But that is a mistake! More elaborate and specific prompts work better.
Don’t ask it to write an essay about how human error causes catastrophes. The AI will come up with a boring and straightforward piece that does the minimum possible to satisfy your simple demand. Instead, remember you are the expert and the AI is a tool to help you write. You should push it in the direction you want. For example, provide clear bullet points to your argument: write an essay with the following points: -Humans are prone to error -Most errors are not that important -In complex systems, some errors are catastrophic -Catastrophes cannot be avoided
But even these results are much less interesting than a more complicated prompt: write an essay with the following points. use an academic tone. use at least one clear example. make it concise. write for a well-informed audience. use a style like the New Yorker. make it at least 7 paragraphs. vary the language in each one. end with an ominous note. -Humans are prone to error -Most errors are not that important -In complex systems, some errors are catastrophic -Catastrophes cannot be avoided
Notice that not every part of the prompt is followed exactly - there are six paragraphs, not seven - but that is typical of how generative AI works: you don’t always get what you directly ask for, but you can push towards something unique and interesting by playing with prompts.
So try asking for it to be concise or wordy or detailed, or ask it to be specific or to give examples. Ask it to write in a tone (ominous, academic, straightforward) or to a particular audience (professional, student) or in the style of a particular author or publication (New York Times, tabloid news, academic journal). You are not going to get perfect results, so experimenting (and using the little “regenerate response” button) will help you get to the right place. Over time, you will start to learn the “language” that ChatGPT is using.
Play with memory and length
ChatGPT’s huge advance over other generative AIs is that it has a memory. It keeps track of what you wrote, and appears to remember about 3,000 words worth of data. However, what it retains, and how that memory is used, is often hard unclear.
Sometimes the memory is useful, you can (and should) ask it to revise previous work: change the third paragraph to be more professional or use a different example in the middle and it will provide a revised version. However, sometimes it forgets what you were working on, and you have to remind it. You may, for example, want to tell it revise the third paragraph on the essay on catastrophes so it doesn’t lose track.
Memory can also be a problem. ChatGPT can get “stuck” repeating the same examples or tone throughout many requests. To unstick it, you may need to ask it to give a different example, or to change its tone, or to alter your early prompts. Often, the best option is to give ChatGPT amnesia: start a new chat and try again.
Similarly, you might want to break up your requests to the Chatbot into smaller chunks. Ask it for an introduction, and revise that to get the tone that you want to achieve. Only then should you start asking for additional paragraphs.
And if the results of those longer pieces cut off, simply asking ChatGPT to “continue” is enough to get the rest of the material.
Play with personas and style
You can ask the AI to use specific styles for writing. You will get different results from asking for an academic essay versus a persuasive article versus a blog post versus a corporate memo. The results are often surprising. So you should consider trying several writing types when experimenting with prompts.
Another way to get interesting writing out of ChatGPT is by asking the AI to be someone else. You can have the AI play characters by prompting it to think of itself as a chef, or a novelist, or Plato. This can lead to delightful results. You can start with this list of hundreds of persona suggestions. I used a modified version of the storyteller prompt: “I want you to act as a storyteller. You will come up with entertaining stories that are engaging, imaginative and captivating for the audience. It can be fairy tales, educational stories or any other type of stories which has the potential to capture people’s attention and imagination. You should provide lots of detail and make the story memorable.”
A very different perspective on error and catastrophe!
Remember limitations… and your ideas?
ChatGPT, like all generative AI systems, is a tool. Tools are used by humans to accomplish specific tasks. Thinking of it that way helps unlock its potential, but also avoid its pitfalls.
For example: Don’t ask it for facts that you can’t easily check. Don’t ask it to provide references. Don’t have it do math, or conduct analysis. It will happily fake doing these things for you and the output will mostly likely be wrong. ChatGPT it is far from the all-knowing AI that the movies taught us to expect. But as a tool to jumpstart your own writing, multiply your productivity, and to help overcome the inertia associated with staring at a blank page, it is amazing.
I’d like to conclude by asking you to write in the comments about any tips or prompt-crafting that you are using to make ChatGPT work. Part of what makes the release of this tool so exciting is that we are learning about it collectively, and that we can only understand the system through exploration. So what prompts or experiences do you want to share?