Introduction to Prompts and Prompt Engineering
Download as PPTX, PDF0 likes125 views
Introduction to Prompts and Prompt Engineering
![N-Shot Prompting
•Zero-shot Prompt - No Examples Given:
•“Write a Python function that converts
Fahrenheit to Celsius.”
One-shot Prompt - One Example Given:
•“Here is a temperature conversion
functions … [SAMPLE CODE] … Now, write
a function to convert Fahrenheit to
Celsius following the same style.”](https://image.slidesharecdn.com/promptengineering-250529184423-3f1eec05/85/Introduction-to-Prompts-and-Prompt-Engineering-45-320.jpg)
![N-Shot Prompting
• Few-shot Prompt – 2 to 5 Examples Given:
• “Here are 2-5 temperature conversion
functions … [SAMPLE CODE] … Now, write a
function to convert Fahrenheit to Celsius
following the same style.”
Multi-shot Prompt – Over 5 Examples Given:
• “Here are over five temperature conversion
functions … [SAMPLE CODE] … Now, write a
function to convert Fahrenheit to Celsius
following the same style.”](https://image.slidesharecdn.com/promptengineering-250529184423-3f1eec05/85/Introduction-to-Prompts-and-Prompt-Engineering-46-320.jpg)
Introduction to Prompts and Prompt Engineering
- 2. Contents Introduction to Prompt Engineering Principles of Effective Prompting Prompting Methodologies 10 General Sample Prompts
- 3. Prompt Engineering • Prompt engineering is designing and refining prompts to effectively communicate with GenAI models, improving quality, relevance, and accuracy of their responses.
- 4. Prompt Engineering •Prompt Engineering can … • increase the accuracy and relevance of answers. • help control the tone, style, and format of answers. • enhance the efficiency of answers.
- 5. Prompt Engineering •It is important to note that different GenAI models interpret prompts differently (ChatGPT, DALL·E, Copilot, etc.), so the most important thing to do is try out different queries and different approaches to generating queries, and see which ones work best.
- 6. Prompt Engineering 3Ps of Prompt Engineering Add more Clarity
- 8. Prompt Engineering •Principles of Effective Prompting: •Clarity & Precision •Context & Constraints •Validation & Verification •Repetition & Repetition
- 10. Clarity & Precision •Using clear and specific language to guide GenAI responses. •Let’s say you want an GenAI to generate a Python function to calculate the factorial of a number >>>
- 11. Clarity & Precision •Bad Prompt - Vague & Ambiguous • Write about photosynthesis.
- 12. Clarity & Precision •Good Prompt - Clear & Precise • Explain the process of photosynthesis in plants, including the role of sunlight, chlorophyll, water, and carbon dioxide.
- 14. Context & Constraints •Providing background information or setting boundaries to get relevant answers. •Let’s say you want to generate a function to sort a list of numbers >>>
- 15. Context & Constraints •Bad Prompt - Lacks Context & Constraints • Generate a recipe for pasta.
- 16. Context & Constraints •Good Prompt - Provides Context & Constraints • Generate a vegetarian pasta recipe that serves 2 people, can be cooked in under 30 minutes, and does not use any dairy products. Include a list of ingredients and step-by-step instructions.
- 18. Validation & Verification •Verifying the output of Generative AI (GenAI) is crucial to ensure accuracy, reliability, and ethical integrity. •Here's a structured approach to critically assess AI-generated content >>>
- 19. Validation & Verification: Check the Facts !!! • Verify Key Information: Scrutinize names, dates, statistics, and claims by comparing them with credible sources like academic journals, official reports, or reputable news outlets. • Validate Citations: If the GenAI provides references, ensure they exist and accurately support the content. GenAI models can sometimes fabricate citations or misattribute information. • Use Fact-Checking Tools: Leverage platforms such as FactCheck.org, Snopes, or Google Fact Check Tools to confirm the veracity of the information
- 20. Validation & Verification: Check the Facts !!! •Take the CRAAP Test!!! •Currency •Relevance •Authority •Accuracy •Purpose • https://damiantgordon.com/CheckSheets/CHECKSHEET-CRAAP-Test.pdf
- 21. Validation & Verification •https://damiantgordon.com/CheckSheets/ Again, lots of Checklists on my website, these can be really helpful for some learners to actively and authentically engage with content and models:
- 23. Repetition & Repetition •Repetition can help improve prompts — but it depends on how and why you use it. •Here’s how repetition can enhance prompt effectiveness, especially with GenAI >>>
- 24. Repetition & Repetition •Clarifying Intent •Repeating or rephrasing your request can reinforce what’s important. •Example: •“Give me three bullet points. Just three clear, short bullet points.”
- 25. Repetition & Repetition •Guiding Structure •GenAI responds well to consistent patterns. •Example: •“Write a short story. Make it short, with a clear beginning, middle, and end.”
- 26. Repetition & Repetition •Stressing Constraints •Repetition can highlight limits (like tone, format, or length). •Example: •“Use plain language. No jargon. Really, keep the language plain and simple.”
- 27. Repetition & Repetition •Prompt Tuning • You can repeat an adjusted version of a prompt to refine the GenAI's direction. •First prompt: • “Summarize this article.” •Follow-up: • “Summarize it again, but focus only on the economic implications.”
- 29. Prompt Engineering •Prompting Methodologies: •Iteration & Testing •Role-based Prompts •N-Shot Prompting •Chain-of-Thought Prompting
- 31. Iteration & Testing •Experimenting with different phrasings and formats to achieve the best output. •Let’s say we want to generate a function that checks if a string is a palindrome (reads the same forward and backward) >>>
- 32. Iteration & Testing Attemp t 3 Attemp t 2 Attemp t 1
- 33. Iteration & Testing •Iteration 1 – A bit vague • Summarize this article.
- 34. Iteration & Testing •Iteration 2 – Adding Constraints • Summarize the main argument of the article in 3 bullet points, each no longer than 20 words.
- 35. Iteration & Testing •Iteration 3 – Final Refinement • Summarize the main argument of the article in 3 bullet points, each no longer than 20 words, and avoid repeating quotes from the text.
- 36. Iteration & Testing •Methodology • Start with a basic prompt and analyze the output. • Iterate by adding constraints based on shortcomings. • Test edge cases to verify correctness. • Refine until the GenAI’s output meets all requirements.
- 38. Role-based Prompts • Asking the GenAI to act as a specific expert or persona. • This helps generate more accurate and more context- aware responses. By assigning a role, you guide the GenAI's thought process and ensure it presents responses to the expected expertise level. • For example, role-based prompt for Business Guru >>>
- 39. Role-based Prompts •Prompt Without a Role - Generic Output • What are some business tips?
- 40. Role-based Prompts •Role-Based Prompt - More Focused & Useful • You are a seasoned startup advisor with 15 years of experience in SaaS businesses. What are the top 5 tips you'd give a first-time founder launching a B2B SaaS product?
- 41. N-Shot Prompting
- 42. N-Shot Prompting •This type of prompting is concerned with whether or not the prompt includes an example …
- 43. N-Shot Prompting Prompt Type Explanation Zero-shot Prompting (0) The GenAI is given no examples and must infer the pattern or solution. One-shot Prompting (1) The GenAI is given one example to learn from before generating an answer. Few-shot Prompting (2-5) The GenAI is given two to five examples to learn from before generating an answer. Multi-shot Prompting (5+) The GenAI is given over five examples to learn from before generating an answer.
- 44. N-Shot Prompting •Why is giving examples better? • More structured & consistent. • GenAI follows the given style. • Consistent formatting and readability. • Encourages GenAI to infer patterns from examples.
- 45. N-Shot Prompting •Zero-shot Prompt - No Examples Given: •“Write a Python function that converts Fahrenheit to Celsius.” One-shot Prompt - One Example Given: •“Here is a temperature conversion functions … [SAMPLE CODE] … Now, write a function to convert Fahrenheit to Celsius following the same style.”
- 46. N-Shot Prompting • Few-shot Prompt – 2 to 5 Examples Given: • “Here are 2-5 temperature conversion functions … [SAMPLE CODE] … Now, write a function to convert Fahrenheit to Celsius following the same style.” Multi-shot Prompt – Over 5 Examples Given: • “Here are over five temperature conversion functions … [SAMPLE CODE] … Now, write a function to convert Fahrenheit to Celsius following the same style.”
- 47. N-Shot Prompting • When to use each: •Zero-shot: Useful for general tasks but may produce inconsistent results. •One-shot: Useful for specific task where the answer needs to closely match the example.
- 48. N-Shot Prompting • When to use each: • Few-shot: Helps guide the GenAI by showing 2-5 examples, improving accuracy and structure. So, use few- shot prompting when you need consistency, style matching, or complex logic. • Multi-shot: Helps guide the GenAI by showing over 5 examples, improving accuracy and structure. So, use multi-shot prompting when you need a very high level of consistency, style matching, or complex logic.
- 49. N-Shot Prompting • NOTE: • With Few-Shot and Multi-shot prompting, you must always be aware of BIAS, there may be something in the specific examples that the GenAI might think have to always be contained in their answers, so review its answers for this.
- 51. Chain-of-Thought Prompting • Also called: • Step-by-Step Instructions • Step-by-Step Prompting • Multi-Turn Prompting • Multi-Step Prompting • Chain-of-Reasoning Prompting
- 52. Chain-of-Thought Prompting •Encouraging step-by-step reasoning for complex tasks. •This helps improve accuracy and ensures logical consistency, especially for mathematical or algorithmic tasks. •For example, we have a function that finds the sum of all even numbers in a list, but it’s giving incorrect results >>>
- 53. Chain-of-Thought Prompting •Initial Prompt •If there are 3 red balls and 5 blue balls in a bag, and you take out 2 balls without looking, what is the probability that both are red?
- 54. Chain-of-Thought Prompting •Chain-of-Thought Prompt •If there are 3 red balls and 5 blue balls in a bag, and you take out 2 balls without looking, what is the probability that both are red? Please explain your reasoning step by step.
- 55. Chain-of-Thought Prompting •Chain-of-Thought Prompt • There are a total of 3 red + 5 blue = 8 balls. • The probability that the first ball drawn is red is 3/8. • If the first ball is red, there are now 2 red and 7 total balls left. • So the probability that the second ball is red is 2/7. • Therefore, the probability that both balls are red is (3/8) × (2/7) = 6/56 = 3/28.
- 56. Chain-of-Thought Prompting •Why better? • Clear breakdown of the problem before answering it. • Step-by-step reasoning helps with learning and debugging. • Reduces GenAI hallucinations by enforcing structured logical thinking.
- 58. Summarization Prompt • “Summarize the key findings of this research paper in three bullet points, focusing on the methodology, results, and conclusions.” • Why it's good: • It specifies the desired format (three bullet points). • It defines the scope (methodology, results, and conclusions). • It ensures clarity by requesting a summary rather than a full analysis.
- 59. Analytical Prompt • “Compare and contrast the strengths and weaknesses of supervised and unsupervised learning, providing examples of real-world applications for each.” • Why it's good: • It requires critical thinking rather than just factual recall. • It demands structured reasoning (compare and contrast). • It provides a clear context (supervised vs. unsupervised learning).
- 60. Creative Writing Prompt • “Write a short science fiction story (500 words) set in a world where AI has become the ruling class. Include a protagonist who challenges the system.” • Why it's good: • It provides a clear theme (AI as rulers). • It specifies word count (500 words). • It introduces a conflict (protagonist challenges the system).
- 61. Code Generation Prompt • “Write a Python function that takes a list of numbers and returns the mean, median, and mode. Ensure the function is efficient and handles edge cases like empty lists.” • Why it's good: • It defines inputs and outputs (list of nums mean, median, → mode). • It includes performance expectations (efficient). • It asks for error handling (empty lists).
- 62. Educational Explanation Prompt • “Explain the concept of entropy in thermodynamics in simple terms, using an everyday analogy a 12-year-old can understand.” • Why it's good: • It specifies the complexity level (simple terms). • It requires an analogy (everyday example). • It identifies the target audience (12-year-old).
- 63. Debate and Argumentation Prompt • “Argue for and against the use of AI in education. Provide at least two strong points for each side, then conclude with your own reasoned opinion.” • Why it's good: • It requires balanced reasoning (both sides of the argument). • It demands specific points (two per side). • It includes a personal conclusion (opinion).
- 64. Data Analysis Prompt • “Given a CSV file containing sales data with columns for date, product, and revenue, write a Python script to calculate the total revenue per product and visualize it using a bar chart” • Why it's good: • It defines the input format (CSV with specific columns). • It describes expected outputs (total revenue per product). • It requires a visualization (bar chart).
- 65. Ethical Discussion Prompt • “What are the ethical concerns surrounding deepfake technology? Discuss both potential harms and possible benefits, citing real-world examples.” • Why it's good: • It prompts critical thinking (ethical concerns). • It requires both pros and cons (harms + benefits). • It encourages real-world applications (examples).
- 66. Roleplaying/Simulation Prompt • “Imagine you are a data scientist presenting findings to a non-technical executive team. Explain the key insights from your predictive model in a way that emphasizes business impact.” • Why it's good: • It provides a clear role (data scientist). • It sets a target audience (non-technical executives). • It emphasizes practical application (business impact).
- 67. Multimodal Prompt • “Describe what a violin sounds like using only visual and tactile metaphors.” • Why it's good: • It forces creative thinking (visual and tactile metaphors). • It removes direct sensory descriptions, making it more challenging. • It encourages unique, engaging responses.