💬 Top Tips for Communicating with LLMs from Google

Google has released a detailed 68-page guide that explains in simple terms how to effectively interact with large language models to get the most efficient and useful responses.

Here are the most valuable takeaways:

1️⃣ Use examples. A prompt without examples (zero-shot) works well only for simple questions. If the task is more complex—like writing an email or a post—add one or more examples (one-shot / few-shot) so the model can understand the task's format, style, and essence.

2️⃣ For complex tasks, use a Chain of Thought. Add this phrase to your prompt: "Let's think step by step."

3️⃣ Provide clear and concise instructions. If a prompt is hard for a human to understand, it will also confuse the LLM.

Example:
❌ "I'm currently in New York, and I'd like to learn more about the attractions. I'm traveling with two three-year-old kids. Where should we go during our vacation?"
✅ "Act as a tour guide for tourists. Describe great places to visit in Manhattan with a three-year-old child."

4️⃣ Control response length. Specify the desired length of the response to get an appropriate answer. For example: "Explain quantum physics in a way that fits into a single tweet."

5️⃣ Take a step back. Start by asking for general ideas—this activates the model's knowledge. Then, use those ideas as context so no details are missed.

Example:
Step 1: "Based on popular first-person shooter games, name 5 fictional key locations that would make a level's storyline engaging and rich."
Step 2: "Using these themes, write one paragraph of a storyline for the location 'Abandoned Military Base.'"

6️⃣ Experiment with styles. The same query can be phrased as a question, statement, or instruction—and yield different results. This is especially useful when creating creative texts.

For example, to gather information for a text about the Sega Dreamcast, you can ask:

• "Why was this console revolutionary?"
• "Write a paragraph about the Sega Dreamcast and its features."
• Or start with: "Sega released the console in 1999. It..."

More useful resources:

📌 Interesting lectures on LLMs and machine learning

📌 Top 5 free AI services for summarizing YouTube videos

#top #education #Google @hiaimediaen