Summary

  • A Google engineer has published a whitepaper on prompt engineering for LLMs
  • The paper offers tips for crafting better prompts to get better responses from AI chatbots
  • Here are five recommendations from the paper for everyday users:
  1. Keep it simple and to the point, focusing on action verbs and avoiding extra words or complex language that could be misinterpreted
  2. Give the AI a role or personality first, such as a travel guide or a kindergarten teacher, to help it adopt a appropriate tone and vocabulary and avoid going off-topic
  3. Provide examples when possible to make the output more focused and structured
  4. Try step-back prompting by asking a broad question first and then feeding the answer back as part of the next prompt to receive more detailed and thoughtful responses
  5. Focus on instructions rather than restrictions, inspiring creativity in the LLM and avoiding guessing about what is allowed
  • These tips can improve the specificity, creativity, and relevance of AI chatbot responses.

By Karrar Haider

Original Article