Blog: Learn Machine Learning from a Google AI Engineer
The Art of Writing Effective Prompts
Whether you are writing a prompt to let the LLM or AI answer a question from a contract, summarize a report, classify user feedback, or extract order information, writing a prompt containing just one line with the question is likely insufficient. You will need to provide instructions, constraints, reasoning, and a few-shot examples, and you will set your expectations for the style and output format. Trust me, you will quickly end up with an extensive written prompt, which increases in size over the various iterations.
Writing prompts can present challenges, particularly when instructions or constraints clash. For instance, an instruction to ‘avoid hallucinations or fabrications’ could hinder creativity or explanations in certain edge cases. This underscores the importance of adaptability in your prompt writing approach. When working with a test set of example user inputs, it’s crucial to be flexible and ready to revisit and verify if recent prompt changes don’t disrupt previously successful prompts. This adaptability is key to maintaining the effectiveness of your prompts.
This is the 5th blog of the series: Prompt engineering for business applications. Prompt Engineering is complex and requires careful planning and refinement to achieve desired results from AI models. As a software engineer @Google with experience in prompt engineering for major businesses, I will share practical learnings in a blog series to help others unlock the power of AI beyond simple tasks.
Whether you are writing a prompt to let the LLM or AI answer a question from a contract, summarize a report, classify user feedback, or extract order information, writing a prompt containing just one line with the question is likely insufficient. You will need to provide instructions, constraints, reasoning, and a few-shot examples, and you will set your expectations for the style and output format. Trust me, you will quickly end up with an extensive written prompt, which increases in size over the various iterations.
Writing prompts can present challenges, particularly when instructions or constraints clash. For instance, an instruction to ‘avoid hallucinations or fabrications’ could hinder creativity or explanations in certain edge cases. This underscores the importance of adaptability in your prompt writing approach. When working with a test set of example user inputs, it’s crucial to be flexible and ready to revisit and verify if recent prompt changes don’t disrupt previously successful prompts. This adaptability is key to maintaining the effectiveness of your prompts.
This is the 5th blog of the series: Prompt engineering for business applications. Prompt Engineering is complex and requires careful planning and refinement to achieve desired results from AI models. As a software engineer @Google with experience in prompt engineering for major businesses, I will share practical learnings in a blog series to help others unlock the power of AI beyond simple tasks.
Lee Boonstra
Lee Boonstra (they/them) has been a presence in the tech world since 2007, wearing many hats from AI software engineer to prompt engineer, web developer to technical trainer, and developer advocate.
With eight years of experience at Google (and AI) under their belt, they now hold the role of AI Software Engineer (SWE) at the Google Cloud office of the CTO. Leading innovation projects, Lee aims to disrupt markets and foster collaboration globally. Their expertise in Conversational and Voice technology, alongside (Generative) AI, has led to recognition as a respected public keynote speaker and published author for O’Reilly and Apress. Lee eases tech headaches and celebrates those light bulb moments.
Lee has written the Google/Kaggle Prompt Engineering whitepaper and two books: "Hands-on Sencha Touch 2" (O'Reilly) and their more recent work, "The Definitive Guide to Conversational AI with Dialogflow and Google Cloud" (Apress).
Documenting Your Prompts a Best Pract...
Beyond the Basics How to Choose and C...
The Foundation Understanding LLMs and...
