Do you Prompt? My AI-LLM Prompt Method

How do you talk to AI?

To interact with an AI, typically through natural language processing and dialogue systems, you can use voice commands (with Text to Speech) or type in written texts, depending on the platform or application. You may need to ask questions, give instructions, or engage in conversation just as you would with another person.

What is an AI Prompt?

An AI prompt refers to a carefully crafted input statement given to an artificial intelligence (AI) system or application, such as a machine learning algorithm or natural language processing model. This statement serves as the starting point for generating responses, predictions, or actions, based on the provided text. The goal of an effective prompt is to ensure that the AI understands and interprets the intended context accurately, while producing relevant outputs. In summary, it's the initial question or instruction used by an AI to process data and provide results.

Why are prompts used?

Prompts are used in various AI applications because they help direct the focus and understanding of the underlying algorithms. They serve as guides for the models to comprehend specific tasks, providing clear instructions about what information should be processed, analyzed, or generated next. Essentially, prompts facilitate communication between humans and machines, enabling efficient interactions and enhancing AI performance within complex environments.

What is Natural Language Processing (NLP)?

Natural Language Processing (NLP) refers to the subfield of Artificial Intelligence that deals with teaching computers to understand human languages, including speech recognition, text analysis, and sentiment analysis. It aims to bridge the gap between how people communicate using their native languages and computer systems' ability to interpret those communications effectively. NLP encompasses techniques like part-of-speech tagging, named entity recognition, parsing, and semantic role labeling among others.

How does Generative AI process a Prompt?

Generative AI processes a prompt by first analyzing its linguistic structure, identifying key elements such as keywords, contextual clues, and grammatical rules. Then, drawing upon vast amounts of previously learned knowledge from training datasets, it generates a response, prediction, or action according to the given prompt. This process involves applying neural networks, statistical modeling, and other advanced machine learning methods to interpret and respond to natural language inputs efficiently.

What is Prompt Design?

Prompt design involves creating well-crafted input statements that enable artificial intelligence systems, particularly in areas such as natural language processing and generative AI, to understand and produce accurate, meaningful, and useful responses. By focusing on essential aspects of context, meaning, and syntax, prompt designers aim to improve the efficiency and effectiveness of these intelligent systems in carrying out complex tasks involving human communication and interaction.

What is Prompt Engineering?

Prompt engineering builds upon Prompt Design by solving a complex user requirement or use-case by factoring in the best prompt, data models, instruction templates, platform, etc… to provide the most optimal output.? Prompt Engineering can involve fleshing of user requirements, design, iterative testing, and analysis to deliver the best results.

When do you use Prompt Design versus Prompt Engineering?

While both prompt design and prompt engineering share similarities in the creation of effective input statements for AI systems, they are distinct practices in terms of scope and purpose. Prompt design focuses more on the development of high-quality prompts that allow the AI to understand and respond to specific situations or tasks effectively. Prompt engineering, on the other hand, takes this concept further by incorporating additional techniques and approaches to optimize the entire process, from selecting the right architecture and training datasets to fine-tuning hyper-parameters and evaluating performance metrics. In summary, prompt design is primarily concerned with prompt quality, whereas prompt engineering combines prompt design with broader optimization efforts throughout the AI pipeline.

What is a good format for a prompt?

I have spent hundreds of hours prompting and testing various methods with LLM(s).? Over this time, I developed my own method of prompting that has worked very well.? I’ve used it to write over 1000 pages of content for books/novels and use it regularly to help perform tasks for my job.? For instance, drafting analysis papers, creating strategic plans, and even creating documentation for a customized tabletop game, that I can play with my kids.

The method I use can be referred to as “PAIR”.?

PAIR Method of prompting:

Context – Set the context for the list of questions you plan to ask, by asking some preparatory questions in advance.? For instance, if you want to put together a solid response on a topic about AI Prompting, you will ask a series of questions, like I’ve laid out in this article.? Logical questions, like “what is AI prompting?”, “What is it used for?”, “How do you use it?”.? Once you have a solid set of answers and content, then you ask the actual question you are trying to answer.? Because LLM(s) process queries in the aggregate (meaning they submit the entire chat history with each new query), the more you ask upfront, the better the accuracy will be.?

Persona - Define a persona for the AI to adopt in preparing to formulate a response.

Action - Define what action you want performed by the AI.

Information - Provide data, information, descriptions, etc... To inform the Action being taken in your prompt.

Response- Define the output and any special formatting, editing, sorting, etc... that you want taken on the response.

??

Examples:

(P) You are an award-winning book editor.? (A) Review the text, correct any grammar mistakes, rewrite any overly wordy or complex sentences, and ensure dialogue for characters is punchy.? (I) This story is meant to be very action packed with a fast pace.? The characters are meant to be direct and not verbose or too talkative. The overall themes of the story should be love, loss, and overcoming grief. (R) Format the response in the form of a third person narrative and ensure dialogue follows the standard for novel writing.

?

(P) You are a bestselling novelist from the south with a flair for the dramatic.? (A) Write 500 words describing someone seeing frost on the kudzu for the first time.? (I) This should take place in the heart of Alabama near the coast.? (R) Format the response in the form of a first person introspective.

?

(P) You are an award-winning tabletop game developer. (A) Generate a player character and associated stats. (I) Each character should include, the Character Name, Character description, Player Faction (Luna, Nightfall, Celestial, Starshine, or Cosmos), Stats (Combat Prowess, Defensive Ability, Intellect, and Situational Awareness), Health (Range 40 to 100), and a random Player Faction bonus. The Faction Bonus should be directly assigned to one of the defined factions. Stats for Combat Prowess, Defensive Ability, Intellect and Situational Awareness, should range from 8 to 18. (R) Format the response as a table with field names and the appropriate stats.

?

Now this method may not work best for every use-case, but I’ve found that it provides the best return for the least amount of effort across a broad range of topics.?

I have also developed my own prompt guide for writing long-format novels, books, etc… that I plan to post about in the future.?

If you’re out there prompting, please give this method a try and let me know how it works out.? I mostly use Textgen-WebUI, a LocalLLM solution but in theory it should work with any SaaS or other LLM platform.