Definition from Webster’s 1913
“Prompt (prŏmt; 215), a. [Compar. Prompter (?); superl. Promptest.] [F. prompt, L. promptus, properly, brought forth (to light or view), hence, visible, evident, at hand, ready, quick, — p. p. of promere to take or bring forth; pro forth + emere to take. See Redeem. ]
1. Ready and quick to act as occasion demands; meeting requirements readily; not slow, dilatory, or hesitating in decision or action; responding on the instant; immediate; as, prompt in obedience or compliance; — said of persons.
Very discerning and prompt in giving orders.
Clarendon.
Tell him I am prompt
To lay my crown at’s feet.
Shak.
And you, perhaps, too prompt in your replies.
Dryden.
2. Done or rendered quickly, readily, or immediately; given without delay or hesitation; — said of conduct; as, prompt assistance.
When Washington heard the voice of his country in distress,
his obedience was prompt.
Ames.
3. Easy; unobstructed. [Obs.]
The reception of the light into the body of the building was very prompt.
Sir H. Wotton.
Syn. — Ready; expeditious; quick; agile; alert; brisk; nimble. — Prompt, Ready, Expeditious. One who is ready is prepared to act at the moment. One who is prompt acts at the moment. One who is expeditious carries through an undertaking with constant promptness.
Prompt, n. (Com.) A limit of time given for payment of an account for produce purchased, this limit varying with different goods. See Prompt-note.
To cover any probable difference of price which might arise before the expiration of the prompt, which for this article [tea] is three months.
J. S. Mill.
Prompt, v. t. [imp. & p. p. Prompted; p. pr. & vb. n. Prompting.]
1. To assist or induce the action of; to move to action; to instigate; to incite.
God first . . . prompted on the infirmities of the infant world by temporal prosperity.
Jer. Taylor.
2. To suggest; to dictate.
And whispering angles prompt her golden dreams.
Pope.
3. To remind, as an actor or an orator, of words or topics forgotten.
From act two, scene two of “an excellent conceited tragedie of Romeo and Juliet”
“Juliet:
By whose direction found’st thou out this place?
Romeo:
By love, that first did prompt me to inquire.
He lent me counsel, and I lent him eyes.
I am no pilot, yet wert thou as far
As that vast shore washed with the farthest sea,
I would adventure for such merchandise.”
Machine Section One
(relevant words, key phrases, and inputs related to large language model prompts)
Generative Model
Pretraining Data
Fine-tuning
Natural Language Processing
Text Summarization
Machine Learning
Artificial Intelligence
Neural Network
Embedding
Transfer Learning
Input Sequence
Batch Size
Tokenization
Positive Reinforcement
Negative Reinforcement
Supervised Learning
Unsupervised Learning Key Phrases:
Large Language Models
Contextualized Encodings
Embedding Layers
Input embeddings
Output embeddings
Attention Mechanisms
Cross-modal Prediction
Language Generation
Language Understanding
Conditional Generation
Distributional Semantics
Conversational AI
Open Domain Question Answering
Multi-task Learning Inputs:
Prompts or Questions
Corresponding Responses or Answers…
Reference Guide: Creating Effective Prompts for Large Language Models
📝 Introduction
Effective prompts are key to harnessing the power of large language models (LLMs) like GPT-3 and GPT-4.
This reference guide provides step-by-step examples and additional resources to help users craft prompts that yield high-quality responses while avoiding common pitfalls.
🌐 Excellent Prompts
Specific Questions: Begin with clear and specific questions, such as “What are the environmental impacts of renewable energy?” to get precise answers.
Contextual Prompts: Provide relevant context before asking questions. For example, “In the context of climate change, explain the importance of reducing carbon emissions.”
Creative Storytelling: Kickstart creative storytelling with an engaging sentence like, “Imagine a world where humans can teleport at will.”
🚫 Prompts to Avoid
Vague Prompts: Steer clear of vague questions that lead to unclear or irrelevant responses.
Biased Prompts: Be mindful of biased prompts that may result in biased output. Avoid questions that assume a particular viewpoint.
(take it further with https://promptperfect.jina.ai/)
🌐 Prompt Databases
ChatGPT Prompts Repository: Explore a wide range of user-contributed prompts for ChatGPT on GitHub
https://github.com/alexkgwyn/chatgpt-prompts
Prompt Library: Access an extensive library of prompts for various AI models, including GPT-3, GPT-4, and more https://github.com/transformers/prompt-library
📖 Additional Websites
Zapier’s Guide: Learn how to write effective prompts for GPT-3 and GPT-4 with Zapier’s guide <a href=”https://zapier.com/blog/gpt-prompt/
Beebom’s Prompt List: Find over 125 ChatGPT prompts for various workflows on Beebom https://beebom.com/best-chatgpt-prompts/
📝 Notes: Crafting Effective Prompts
Creating effective prompts is both an art and a science. To achieve the best results, follow these steps:
Define Your Goal: Clearly define the desired output or information you seek.
Provide Context: Start with context to help the model understand the task or question.
📖 Extra Glossary
GIGO Factor: “Garbage In, Garbage Out” principle, emphasizing that the quality of prompts impacts the quality of AI-generated output.
🌐 A couple of resources
https://www.zdnet.com/article/how-to-write-better-chatgpt-prompts ZDNet — How to write better ChatGPT prompts
https://zapier.com/blog/gpt-prompt Zapier — How to write an effective GPT-3 or GPT-4 prompt
Large Language Model Prompts List
available resources and links for LLM prompts:
https://huggingface.co/docs/transformers/examples/prompt_learning
Hugging Face Transformers Documentation</a> — This documentation provides a comprehensive overview of prompt learning, including how to create and use prompts for LLMs.
https://github.com/google/prompt-engineering
Google Prompt Engineering Repository — This repository contains code and resources for prompt engineering, including a list of pre-defined prompts for LLMs.
https://aihub.cloud/blog/2023-09-05/prompt-engineering-for-large-language-models/
Prompt Engineering for Large Language Models — This blog post provides a high-level overview of prompt engineering, including best practices and examples.
Machine Section Two
Instructions for creating effective prompts for LLMs:
Be clear and specific in your instructions.
Provide the LLM with as much context as possible.
Use examples to illustrate what you are looking for.
Break down complex tasks into smaller, more manageable steps.
Possible innovations in the future of LLM prompts:
**Interactive prompts:** Prompts that can be updated and refined in real time, based on the LLM’s output.
**Context-aware prompts:** Prompts that can automatically adapt to the context of the task at hand.
**Multimodal prompts:** Prompts that incorporate multiple types of data, such as text, images, and audio.
**Additional Resources:**
* **PromptBase:** A collection of over 10,000 prompts for a variety of tasks, including creative writing, code generation, and translation.
* **PromptHub:** A website where users can share and discover prompts for LLMs.
* **PromptSource:** A collection of prompts for specific LLM models, such as GPT-3 and Bard.
* **PromptLearning:** A website that provides resources for learning about prompt engineering and prompt learning.
* **PromptEngineering:** A blog that covers the latest research and trends in prompt engineering.
* **Prompt Engineering for Large Language Models** by Google AI
* **How to Write Prompts for Large Language Models** by Hugging Face
* **A Guide to Prompt Engineering for Large Language Models** by PromptBase
EPILOGUE
Correcting Grammar and Spelling Errors;
Explicit Correction:
If you want the model to correct grammar or spelling errors in the input text, you can include explicit instructions like, “Please correct any grammar or spelling mistakes in the following sentence: ‘He run fastly.’”
Contextual Correction:
Alternatively, you can frame the request within a contextual prompt, such as, “In this passage, there may be some grammatical errors. Please provide a revised version with correct grammar and spelling.”
Refraining from Translation:
Clear Instruction:
If you don’t want the model to translate text, specify this clearly in your prompt. For example, “Do not translate the following text; instead, provide an analysis.”
Language Detection:
You can add a step to detect the language of the input text and then instruct the model accordingly. For instance, “First, identify the languages used in the following text, and then provide an analysis without translation.”
Handling Mixed Languages:
Contextual Clarification:
If you encounter mixed languages in the input, clarify the context and intent. For example, “The following message is in a mix of English and French. Provide a summary of the English parts.”
Language Separation:
You can specify which parts of the mixed-language text you want to focus on. For instance, “Analyze the French sentences in the following text while ignoring the English portions.”
Politeness and Tone:
Explicit Instruction:
To control the tone and politeness, include explicit instructions like, “Please generate a polite response to the following message.”
Sample Responses;
Provide examples of the desired tone, such as “Respond to the following message with a friendly and helpful tone, similar to ‘Hi, how can I assist you today?’”
Avoiding Bias and Controversial Content:
Guidance:
Clearly instruct the model to avoid generating biased or controversial content. For instance, “Do not generate responses that promote any form of bias or controversy.”
Review Output:
Always review the generated output to ensure it aligns with ethical and responsible guidelines.
Multiple Steps and Context:
For complex tasks, break them down into multiple steps and provide context gradually. Begin with a high-level request and refine it based on the model’s initial response.
Experiment and Refine:
Experiment with different prompts and iteratively refine them to achieve the desired output. Test and adjust prompts as needed to improve results.
Evaluate Output:
Always evaluate the model’s output critically to ensure it aligns with your requirements and objectives. If the response is incorrect or inadequate, consider adjusting your prompt or providing more context.
Remember that prompt design may require experimentation and fine-tuning to achieve optimal results. Clear and specific instructions are key to guiding the model effectively, and iterative testing can help improve the quality of responses. Additionally, ethical considerations should always be at the forefront when crafting prompts to ensure responsible AI use.
____________________________________________________________
mail:
inkrealm@inkrealm.info
Tél :
1 646 801 97 13