Director, Learning Design & Development| PMIAsheville, NC, United States
With Generative AI, iteratively refining and optimizing prompts can lead to better AI-generated results. This may involve adjusting the specificity or clarity of the prompt to increase relevance and accuracy of results.
What examples do you have of how improving a prompt drastically changed the output quality? What specific changes did you make that led to the improvement?
Solomon OkhifohTechnology Project Manager| City of EdmontonGrande Prairie, Alberta, Canada
When generating prompts, clarity and precision are crucial, similar to giving instructions to a person. I have often encountered situations where GenAI, when prompted, provides generic and unusable information. Refining the prompt, just like rephrasing a sentence to a human, has significantly changed the results. For example, if I tell my son, "Go to the room and grab a pair of shoes for me," without specifying the exact pair, he will choose any shoes. However, if I am specific by saying, "Grab the pair of black shoes on top of the rack," he knows exactly what to get. Thus, precise prompts yield precise outputs. Saving Changes...
Q: What examples do you have of how improving a prompt drastically changed the output quality? What specific changes did you make that led to the improvement?
A: Recently, I was comparing FIDIC contracts book (those who know, they are known by color, RED, YeLLOW, SOLVER, GOLD and so on. I wanted Infinity to compare them for certain clauses like 14.0 about "Payments" and so on. My first prompt was unanswered and second was close to analysis. Refining, I asked for "similarities and differences" among the books, then Infinity gave me precise output. Bravo! Saving Changes...
Refining a prompt can significantly impact the quality of output when working with Generative AI (GenAI) models. Here are a few key observations and examples from my experience:
1. Precision in Language Improves Relevance
Example: When generating text summaries or reports, an initial prompt like, “Summarize the latest advancements in AI,” might yield a broad and somewhat unfocused response. Refining it to, “Summarize the latest advancements in AI with a focus on natural language processing (NLP) techniques from 2023,” produces a far more relevant and specific output.
Impact: By including additional context and specifying constraints, the AI is better guided, and the output becomes more aligned with the intended goal.
2. Clarifying the Desired Output Format
Example: When requesting technical documentation or content structured in a particular way, a prompt like, “Write an article about machine learning,” may not result in an organized piece. Refined to, “Write a 5-paragraph article about machine learning, including an introduction, three key techniques, and a conclusion,” the output becomes structured and easier to use.
Impact: Clear instructions on format ensure that the results are immediately useful without needing significant editing or restructuring.
3. Using Iterative Prompt Engineering for Specificity
Scenario: In a data analysis context, if an AI model is used to generate Python code, a prompt like, “Generate a Python script for data analysis,” might produce a simple or generic script. However, refining it to, “Generate a Python script that loads a CSV file, cleans missing data, and visualizes key statistics using Matplotlib,” results in a detailed, relevant script that addresses the specific needs of the analysis.
Impact: Refinement can transform vague and less actionable outputs into highly relevant and actionable content.
4. Addressing Ambiguity Reduces Misinterpretation
Example: In creative writing tasks, a prompt like, “Write a story about a magic forest,” may yield an engaging narrative but lack certain themes or tones. By specifying, “Write a suspenseful story about a magic forest where the protagonist discovers a hidden curse,” the output not only becomes richer in detail but also aligned with a desired mood.
Impact: Reducing ambiguity helps the AI understand the exact creative direction, enhancing the quality of the narrative.
5. Incorporating Examples or Constraints
Example: When generating marketing copy, a prompt like, “Create a tagline for a new electric vehicle,” may generate a variety of options, but adding constraints such as, “Create a short and catchy tagline for a new electric vehicle emphasizing eco-friendliness and innovation,” can refine the suggestions to be more aligned with brand messaging.
Impact: Including examples or constraints ensures outputs are not only creative but also align with specific marketing goals or brand guidelines.
Key Takeaway
Refining a prompt by adding context, specifying format, reducing ambiguity, and including constraints can turn a generic and unhelpful response into a high-quality and purpose-driven output. It’s an iterative process that often requires testing and adjusting the prompt multiple times to get the best results.
Refining prompts resulted in more informative output projecting elaborate detail that's relevant and concise, especially when broken down by prompt chain. Saving Changes...
Refining the prompt greatly improves results. It reminds me of the kids assignment where they are meant to give instructions to make a pb&j sandwich and the person follows every step literally even if it does not make logical sense. GenAI is similar. You can't assume or imply anything in your instructions if you want them to work to create the end product...or sandwich. Saving Changes...
In my experience, refining a prompt can make a huge difference. Even small tweaks often lead to much more accurate and relevant responses, bringing the output much closer to what I was looking for.
For example, when drafting product descriptions for email newsletters, my first prompt was something like, “Write a product description for a summer journal.” The result was plain and didn’t feel engaging. After refining it to, “Write a lively, descriptive product intro for a summer-themed journal that highlights its versatility for travel and everyday use,” the response was much more vibrant and captured the journal’s appeal perfectly for the email.
You can save time, altough you also need to double check information, it's worthy time invested so no doubt using GenAI improves your productivity Saving Changes...
You can save time, altough you also need to double check information, it's worthy time invested so no doubt using GenAI improves your productivity Saving Changes...
Stephen MarshallProduction Manager| ASCDAlexandria, Va, United States
Found that using the Prompt Chaining technique has greatly impacted the AIs usefulness toward answering much more specific questions as well as helping to separate the wheat from the chafe when your initial prompt was either to wildly broad or not as specific in nature as it should have been to produce the information that would be the most helpful Saving Changes...
PMI Team Member
