Generative AI feels like it’s everywhere these days, but how do you generate actionable business value from it? The best way to get the results you need from generative artificial intelligence (AI) is to provide the right prompt to the large language model (LLM) model. This takes practice and fine tuning and a little bit of trial and error, but it will be worth the effort in the end.
The art of creating the perfect prompt is called “AI Prompt Engineering.” There are levels of prompt engineering. For example, you might just want to ask a question of a conversational AI Chatbot and get a response or you may want to create input to a LLM model within a process to make automated decisions.
What is an AI prompt?
First, you need to understand the purpose of the AI prompt. A prompt is text that makes a request of generative AI to perform a task. This might be to write a response to a question, provide the best options for traveling, create a video or image, or even to write a poem. The request is processed by very large machine learning models that are pretrained on huge amounts of data.
These machine learning models can do all sorts of things, so if you want something very specific, you need to provide specific instructions to increase the likelihood of obtaining the results you are requesting. Fine-tuning your prompt, or prompt engineering, can help you achieve the desired results.
Let’s try an example. Assume you would like to write a short poem about an otter playing in a stream that has lost his family. He is enjoying the stream and his solitude, but misses his family. Happily, he eventually finds them. You use a generative AI engine and enter the prompt:
Write a short poem about an otter in a stream
The system responds with something like this.
But you realize that you didn’t give the engine information about the fact that the otter was looking for his family. So, you modify your prompt and try this:
Write a short poem about a lonely otter in a stream enjoying his solitude while searching for his family
This time, the system adds in some details about the fact that he is searching, but happy to be alone too because he has hope that he will find his family someday.
You realize that you forgot to indicate that the otter eventually finds his family, so you make another modification. You also decided you want to cap the number of words at less than 250.
Write a poem under 250 words about a lonely otter playing in a stream enjoying his solitude while searching to reunite with his family that he finally finds
Now you are getting closer to your desired results, but make one more attempt with more .
Write a short poem of 250 words or less about an otter playing in a stream that has lost his family. He is enjoying the stream and his solitude, but misses his family. Happily, he eventually finds them.
As you can see in this simple example, providing specific and clear prompts will help you achieve the most accurate results. This is the art of prompt engineering. With the right prompts, you can start to let AI go to work for you in an automated fashion.
Note: The chatbot used in the previous example conversational AI is ChatGPT.
Roles in prompt engineering
Before we dive into how you might take advantage of using chatbots to automate and improve your business, it is important to understand some key vocabulary.
- Prompt engineers
These professionals focus on designing and fine-tuning prompts for AI systems. They employ creativity and experimentation to develop input text that enables the AI to engage with users in a meaningful manner. - Prompt techniques
Prompt engineers utilize various methods, such as chain-of-thought (CoT) prompting, which divides complex reasoning into intermediate steps. These methods are known as prompt techniques. - Prompt content
Each word in a prompt is important, and even a minor change in phrasing can produce a different response. For example, “describe generative AI” versus “explain generative AI” may result in similar, but different responses.
Each role plays an important part in building the proper prompt for the best response. Having the right individuals with the proper skills will make your organization more successive as they embrace AI.
Why is prompt engineering important?
Prompt engineering directly affects the performance and efficiency of the generative AI models, especially when using natural language processing (NLP). With fine-tuning and proper prompts, you can improve accuracy, maximize the potential of the model, conserve time and resources, and improve the user experience.
Prompt engineering can be used to improve results with a chatbot, but can also be embedded or integrated into processes to automate tasks and achieve better outcomes. Stay tuned for our next blog in this series where we examine this technology in more detail and consider how you can best use it in your business.
Generative AI and Camunda
In case you’re wondering, of course, Camunda works with multiple generative AI models, including OpenAI, Gemini, and others, and you can use these prompt engineering techniques to improve the way generative AI enhances your process orchestration and automation. Feel free to experiment with a free trial, and look out for more tips to help you get started in our next post.
Start the discussion at forum.camunda.io