The usage of large language models (LLMs) like ChatGPT is exploding across industries. Even scientists are leaning on AI to jot down or a minimum of polish their work. A recent evaluation of 5 million scientific studies published in 2023 found a rapid increase in certain words, resembling meticulous, intricate, and commendable. The explanation for this increase? The usage of LLMs.
AI’s impact is already massive, rapidly transforming the workplace. But we’re still on the early stages, attempting to strike a balance between harnessing AI without leaning an excessive amount of on it and compromising quality. At my company Jotform, we have now incorporated AI tools to automate tedious tasks, or as I call it, “busywork,” and unlock employees to deal with the meaningful work that only humans can do. I view AI tools like ChatGPT as invaluable resources—low-cost, efficient AI assistants. But as anyone who’s used ChatGPT will inform you, it’s not 100% reliable or accurate. And it’s only as effective because the prompts you give it.
I recently asked ChatGPT learn how to develop your prompt engineering skills. The primary response was: “Experimentation and Iteration: Constantly experiment with several types of prompts and refine them based on the AI’s outputs. Track what works and what doesn’t, and iteratively improve your prompts.”
In other words, practice makes perfect.
With this in mind, listed below are my top strategies for taking your prompt writing to the subsequent level.
Try meta-prompts
Communication is the lynchpin of a successful relationship. Understanding how people communicate and learn how to best communicate with them will help grow a stronger bond and overcome inevitable misunderstandings when there may be a disconnect between words and intentions or emotions. In the identical vein, your interactions with ChatGPT can vastly improve in case you explicitly ask it learn how to best communicate with it. I commonly ask ChatGPT learn how to phrase prompts so as to get the knowledge or feedback I’m in search of. The more information you give ChatGPT in regards to the results you’re after, the higher it may possibly show you how to generate an efficient prompt.
Here’s a sample meta-prompt:
“How would you phrase a prompt to get detailed feedback on an article from ChatGPT? What information would you include to get probably the most relevant insight?”
In seconds, ChatGPT churned out a prompt template with a temporary explanation of why it included each a part of the prompt. This sort of insight helps you write higher prompts in the long run.
Make it an ongoing dialogue
In most research tools, each query is isolated. You possibly can’t construct on a previous search to refine or dig deeper. ChatGPT has the advantage of enabling you to proceed a thread, similar to a dialogue between two people—a Socratic dialogue, in case you will, where you might be the teacher and ChatGPT is the scholar. I find this capability especially useful if I’m recent to a subject and would really like to start with a basic level of understanding before diving into the nitty gritty.
To interact in dialogue with ChatGPT, start with a broad prompt and refine the following prompts based on the responses.
A sequence of prompts might seem like this:
“In a couple of sentences, are you able to explain the mechanics of a tennis serve?”
“Why does the server shift their weight from the back foot to the front foot?”
“Are you able to briefly explain kinetic energy transfer?”
This method can show you how to to get more nuanced responses and explore areas and perspectives you may not have otherwise considered.
Give ChatGPT a reference text
ChatGPT is usually a little bit of a smooth talker. Ask it about any subject under the sun and it’ll happily wax poetic for so long as you’d like. The rub: it’s not at all times accurate. That’s why providing ChatGPT a reference text will help make sure the replies you receive are grounded in reliable sources.
As OpenAI explains, “In the identical way that a sheet of notes will help a student do higher on a test, providing reference text to those models will help in answering with fewer fabrications.” Feed ChatGPT the sources that you simply already know and trust and let the LLM extract the knowledge you would like.
The prompt may very well be:
“Use the next article in quotes to reply questions.”
ChatGPT: Sure, please go ahead and supply the articles you want me to make use of for answering your questions.
Then, provide ChatGPT with the article (title and publication will often do the trick) and your questions.
Assign ChatGPT a job
One of the vital useful, and admittedly fun, capabilities of ChatGPT is to adopt different personas. You most likely know the parable of the blind men and the elephant. Each man feels a unique a part of the elephant and describes the animal based on their limited perspective. As humans, we are likely to assume our reality is correct even when it’s based on a limited perspective. Incorporating various perspectives empowers us to grasp something more fully.
By simply instructing ChatGPT to adopt a persona, we will get a brand new perspective.
Consider these two prompts and personas:
Prompt: Rewrite the next text from the angle of a poet: A bridge over water connects two lands.
ChatGPT:
Between shores, where rivers dance and whisper,
A bridge, slender and resolute, spans the liquid silver,
Uniting distant lands with an arch of promise,
A testament to the craving of hearts and soil.
Prompt: Rewrite the next text from the angle of an engineer: A bridge over water connects two lands.
ChatGPT
A bridge over water functions as a structural solution, effectively connecting two separate landmasses. It spans the gap with precision-engineered materials, ensuring stability and sturdiness while facilitating transportation and communication between the 2 regions.
At Jotform, our marketing team might use this system to look at a problem from the angle of various buyer personas. It’s an easy technique with myriad use cases.
Model the reply you wish
While ChatGPT won’t be probably the most reliable narrator, it is incredibly logical. Somewhat than attempting to describe the response you wish, which will be surprisingly difficult, you may give ChatGPT an example and it’ll apply the identical logic.
Let’s say I’m working on a marketing campaign and need to make use of ChatGPT to give you snappy, compelling copy.
I could use the next prompt:
Prompt: “I’ll write words and I’d such as you to give you a corresponding adjective. For instance, if I write ‘pretty,’ you may write “more beautiful than the celebrities within the sky.”
ChatGPT:
In fact, I’m ready! Please provide the words you want me to remodel.
Me: nervous
ChatGPT: Nervous as a leaf trembling within the gale of a tempest.
I’d consider that on second thought, that’s not the tone I’m after. So I tweak the prompt.
Prompt: I’ll write words and I’d such as you to give you a corresponding adjective. For instance, if I write ‘pretty,’ you may write “cuter than the prize-winning pig on the Texas State Fair.”
ChatGPT: Got it! Please provide the words you want me to remodel.
Me: nervous
ChatGPT: Nervous as a cat at a rocking chair convention.
On this case, I’d get an idea for my marketing copy—and a chuckle, at no extra charge.
