The [SITE] Framework for LLMs — Part 4: Expanding

Table of Contents

1. [S]ummarizing
2. [I]nferring
3. [T]ransforming
4. [E]xpanding

Introduction

This is Part 4 in a series exploring the potential of Large Language Models (LLMs) and how Product Managers can leverage this technology to drive new features and products. Understanding how to optimize prompts and how to integrate API calls to LLMs becomes an invaluable skill set for any Product Manager leading an AI product.

In a previous post we looked at First Principles of Prompt Engineering. With First Principles covered, we’ll dive deeper into the four primary capabilities of LLMs. Those capabilities are 1) Summarizing 2) Inferring 3) Transforming 4) Expanding.

Knowing how to effectively leverage these capabilities can be a really powerful framework for integrating and leveraging LLMs into your apps. I refer to this as the “SITE” framework for LLMs.

Today’s post focuses on the fourth capability: Expanding.

[E]xpanding

As Product Managers and app developers, we are always looking for innovative solutions to complex problems. One such solution lies in the application of large language models for text expansion. Text expansion is the process of taking a compact piece of text, such as a set of instructions or a topic list, and having a large language model generate a more extensive piece of text, such as an email or an essay.

Brainstorming

Text expansion serves as a powerful brainstorming tool, particularly when using large language models. By feeding the model with a brief prompt, it can generate ideas and content that you might not have considered. This attribute makes text expansion a valuable ally in creative pursuits like content creation, programming, and design.

Enhancing Customer Interaction: A Practical Example

Let’s delve into a practical example of how to use a language model for text expansion. We will generate a personalized email response to a customer review.

We utilize the language model to draft an email based on the customer review and its sentiment. If the sentiment is positive or neutral, we thank them for their review; if negative, we offer an apology and suggest reaching out to customer service.

Suppose you received this review:

Using the prompt below, the language model can create a personalized response that addresses specific details from the review, adding a touch of authenticity and personalization, while maintaining a professional tone.

Dear Valued Customer,

Thank you for taking the time to leave a review about our product. 
We are sorry to hear that you experienced an increase in price and 
that the quality of the product did not meet your expectations. 
We apologize for any inconvenience this may have caused you.

We would like to assure you that we take all feedback seriously and we 
will be sure to pass your comments along to our team. If you have any 
further concerns, please do not hesitate to reach out to our customer 
service team for assistance.

Thank you again for your review and for choosing our product. We hope to 
have the opportunity to serve you better in the future.

Best regards,
AI customer agent

Playing with the Temperature Parameter

One aspect to consider when using large language models for text expansion is the ‘temperature’ parameter. This parameter allows you to control the degree of randomness in the model’s responses, effectively acting as a dial for creativity.

With a temperature of zero, the model opts for the most probable next word, resulting in predictable outputs. In contrast, a higher temperature leads to a more varied response as the model may choose less likely next words.

As tech professionals, you may want to use a lower temperature setting for applications requiring predictable responses, like in customer service bots or factual content generation. But if your task requires more creativity, like brainstorming sessions or content ideation, a higher temperature might be more suitable.

Final Thoughts

In conclusion, text expansion using large language models opens a world of possibilities. However, it also comes with the responsibility of fair use and transparency. As we continue to explore and innovate in this field, understanding how to best leverage and control these tools becomes increasingly critical.

For the full list of articles in this mini series on the [SITE] framework for LLMS check out the posts below:

1. [S]ummarizing
2. [I]nferring
3. [T]ransforming
4. [E]xpanding

If you enjoyed this mini series post please check more in my larger blog series AI For Product People.