4 minute read

Clear communication is critical for effective teams. Using simple, precise language helps bridge gaps across roles. Clear language enables everyone to contribute meaningfully. This is particularly true for technical projects – where complex ideas are shared constantly between domain experts, engineers, and non-technical teammates.

Communicating concepts clearly is a shared responsibility. Relying heavily on jargon can be a barrier; it excludes people from contributing and requires continual translation between groups. It’s mostly just overhead. When team members understand each other, collaboration becomes easier – everyone can contribute to the project.

The Importance of Clear Communication with LLMs

With Large Language Model-driven solutions, clear communication becomes more vital. Non-technical teammates with niche domain knowledge can make a huge impact simply through their written words. With LLMs, written language can become running software. In order to contribute, those teammates need to understand how LLM systems work on their own terms.

We can help our team win by making technical concepts more accessible. We should take steps to learn how to do that. We can’t just learn more about the technology. Will learning about that new vector database improve how your team solves problems? Maybe it will, maybe it wont. How about spending that time learning how to talk about the work in terms more teammates understand?

Simplifying AI Terminology

I found a fantastic resource that embodies this approach: Parlance Labs’ AI communication cheat-sheet. This is part of their free email course, AI Mastery for Executives: A Unique, No-Bullshit Course on AI, which delivers succinct daily tips to bridge the AI communication gap between people who know how to build with the tools, and people who know what problems are worth solving.

The cheat-sheet takes common LLM terms and explains them in simple language that makes sense to everyone. Here are a few examples:

  • Chunking: Breaking long texts into smaller parts.
  • Context Window: The maximum text the model can use at once.
  • Chain of Thought: Prompting the model to think and plan before answering.
  • Inference: Getting an answer back from the model.
  • Retrieval Augmented Generation (RAG): Providing relevant context to the LLM.

Using these simpler terms can make a big difference. For instance, if you mention “RAG” to a group without LLM experience – they might not know how to contribute, and might disengage. But if you say “providing relevant context to the LLM,” they may think, “Hey, I have relevant context for that problem to suggest!” This small shift can open up discussions and invite contributions from all team members.

I’ll be keeping this cheat-sheet on hand to help simpify my own language. I want to get people excited about using LLMs to build solutions and solve real-world problems. Communicating with simple terms helps towards that goal.

Learning how to Write Clearly

If you’re technical and typically use your learning budget on the newest LLM library or technique, why not consider learning about clear communication and writing instead? One of my favorite books on this topic is On Writing Well by William Zinsser. It’s packed with advice on clear writing, making it incredibly relevant for those of us looking to improve how we convey ideas in both technical and non-technical settings.

ChatGPT’s Summary of On Writing Well

With the lens of ‘Inviting AI to the table’ - I used ChatGPT to get a summary of the book. Here’s the prompt:

Prompt: You are an expert in summarizing and distilling the top take-aways from non-fiction books. Your audience is a Software Engineering team that is looking for practical advice on how to improve their writing and communication. They want advice both for communicating with their team, and for writing prompts for Large Language Models.

Your job is to:

  • Research and understand the book ‘On Writing Well’ by William Zinsser
  • Determine the 5 most applicable pieces of advice in the book, that could be useful to Software Engineers
  • Distill this advice into no longer than 3 sentences each
  • Create a bulleted list containing this advice
  • Ensure the advice is written clearly and succinctly

Here’s what ChatGPT generated. It’s pretty good (but you should still read the book):

  • Be Clear and Simple: Avoid jargon, and use straightforward language. Clarity reduces misinterpretation and ensures your message or prompt is understood.
  • Cut Out Clutter: Eliminate unnecessary words. Concise writing is more impactful and easier to interpret, especially for technical communication.
  • Focus on the Reader’s Needs: Write from the reader’s perspective, whether it’s your team or a model. Tailor the message to be accessible and relevant to them.
  • Convey One Idea at a Time: Keep each sentence focused on a single idea to maintain structure and readability. This clarity is crucial for both instructions and prompt crafting.
  • Revise for Precision: Editing is essential. Take time to revise, as small changes can greatly enhance clarity and ensure your instructions are unambiguous.

Investing time in improving how we communicate, especially in writing, is worth it. Whether it’s to empower our team or make our LLMs more effective, clear communication is a foundational building block for building successful teams and successful products.

Updated: