Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Introduction
GPT-3 is powerful, but only if you ask the right way. GPT-3 prompts are the instructions you give to get good outputs. A bad prompt gives bad results. A good prompt gives amazing results. This post teaches you how to write effective prompts.
What Is a Prompt?
A prompt is the text you send to GPT-3. It can be a question, a command, or a few examples. The model reads your prompt and continues the text.
For example:
- Bad prompt: “Write about dogs”
- Good prompt: “Write a 100-word paragraph explaining why golden retrievers make great family pets. Use a friendly, enthusiastic tone.”
For basics of text generation, read text generation.
Prompt Structures That Work
1. Instruction + Context + Format
“You are a travel agent. Recommend a 3-day itinerary for Paris. Include museums, restaurants, and transportation tips. Output as bullet points.”
2. Few-Shot Examples
“Translate English to Spanish:
Good morning → Buenos días
How are you? → ¿Cómo estás?
Thank you very much →”
3. Role Playing
“Act as a career coach. Ask me five questions about my skills and interests. Then suggest three job titles.”
Common Prompt Mistakes
| Mistake | Why It Fails | Fix |
|---|---|---|
| Too vague | GPT-3 guesses poorly | Be specific |
| No format | Output is messy | Specify bullet points, JSON, etc. |
| Too long | Model loses focus | Break into steps |
| Contradictions | Confused output | Keep instructions clear |
Advanced Prompting Techniques
Chain of Thought
Ask GPT-3 to explain its reasoning step by step. This improves accuracy.
Example: “Solve this math problem. Show your work step by step. 24 + 47 = ?”
Temperature Setting
Lower temperature (0.2) = more factual, repetitive. Higher temperature (0.8) = more creative, random.
For API settings, see GPT-3 API tutorial.
Real-World Prompt Examples
Blog post outline
“Create an outline for a 1500-word blog post titled ’10 Tips for Remote Work Productivity.’ Include H2 headings and 3 bullet points per section.”
Code generation
“Write a Python function that takes a list of numbers and returns the average. Include error handling for empty lists.”
Creative writing
“Write a haiku about autumn leaves falling. Use nature imagery and a melancholy tone.”
How to Iterate and Improve
Prompting is trial and error. Start simple. See what GPT-3 gives. Then add more details. Save your best prompts for reuse.
For more on GPT-3’s capabilities, see GPT-3 use cases.
FAQ
1. How long should a prompt be?
Usually 50–300 words. Very long prompts can confuse the model.
2. Can I use emojis in prompts?
Yes. GPT-3 understands emojis.
3. Does prompt order matter?
Yes. Put the most important instructions first.
4. Where can I learn more?
Return to GPT-3 guide or read GPT-3 limitations.
Conclusion
GPT-3 prompts are the key to getting great results. Be specific, provide examples, and specify output format. Experiment and refine. Good prompting saves time and money.
Next: GPT-3 API tutorial or GPT-3 vs GPT-4.
