Hey there,

Joris here.

Life's been a bit of a whirlwind lately, with freelancing taking me on a rollercoaster of tasks. And you know what task always manages to sneak in?

Those meeting summaries. They eat up time like there's no tomorrow.

Our mission today?

To give you more time for the things that truly matter. In this edition, we're delving into a world where AI steps up as your trusty partner, tackling those everyday tasks that can easily bog us down.

Get ready to explore the magic of crafting a speech-to-text assistant that's here to make your life easier.

Today's Unchained Insights

🔍 LangChain Code Snippet: Get AI to Summarize Meetings in Less Than 10 Lines of Code

💡 Project Idea: Elevate Your AI Assistant Game

📚 AI Research: Boosting Your Brainstorms with AI

Introduction

This isn't just about technology; it's about carving out a space to do what truly matters.

Together, let's explore the magic that happens when OpenAI Whisper, GPT3.5, and LangChain come together.

Let's dive in and reclaim some of that precious time.

LangChain Code Snippet

GitHub Repo

Access the complete code and resources in the GitHub Repo.

Don't forget to rename the .env_EXAMPLE file to .env and replace the variable with your personal API Key. I've implemented this intermediate step for security reasons, ensuring your data remains protected.

5 Steps to Your AI Meeting Summaries

1. Import dependencies

2. Load environment variables

3. Transcribe Audio with OpenAI Whisper

4. Configure Large Language Model

5. Summarize Transcript

All wrapped up in just 8 lines of code.

Peek Under the Hood

It’s actually quite simple, here's how it works:

1. Load Environment Variables: Set your OpenAI API Key.

2. Transcribe Audio: Use OpenAI’s Whisper to transcribe your audio file. You can also directly translate the file if you’d like. (Documentation)

3. Initialize Large Language Model

4. Initialize LLMChain:

   1. LLM: Your initialized Large Language Model.

   2. Prompt: The prompt in which the transcript is put.

5. Run the chain: With the LLMChain, voilà—an AI-generated summary of your audio, as if your assistant was taking notes.

Isn't it amazing how just a few lines of code can orchestrate this symphony of AI and bring your assistant to life?

Ready to upgrade your AI Assistant?

Unchained Project Idea: AI Assistant

Looking to supercharge your AI management assistant?

Here's an exciting idea to consider. Imagine an AI-powered ally that not only creates meeting summaries but also boasts additional functionalities to make your workflow even smoother.

This week, let's explore enhancing your assistant with these features:

• Create UI in Streamlit (Documentation)

• Upload audio file (Documentation)

• Handle longer input audio files.

• Zapier integration to send email with summary. (Documentation)

• Zapier integration to create Google Calendar events (Documentation)

Please share your project with me on Twitter and the best one will feature in next week’s AI Unchained.

Unchained Project of The Week

Last week, I set you the challenge of delving into advanced Prompt Generator projects.

Hard Kothari truly outdid himself by making an exceptionally sophisticated prompt improver. This tool is your secret weapon for enhancing your prompt game and getting remarkable results.

What can you achieve with it?

• Effortlessly input your initial prompt

• Select a specific task for the model

• Add special instructions to refine the output

• Generate an improved prompt that's tailored to your needs

• Witness the transformation as your output takes shape

Immerse yourself in this innovation by trying out the project here. If you're as impressed as I am (and I know you will be), spread the word and give Hard Kothari a shoutout on Twitter.

Your appreciation fuels the spirit of AI Unchained and propels us towards even greater heights of creativity.

AI Research

Title: Less Likely Brainstorming: Using Language Models to Generate Alternative Hypotheses.

Short summary: This article introduces a new task called "less likely brainstorming" that aims to generate outputs that are relevant but less likely to happen. The authors propose a controlled text generation method that encourages models to differentiate between generating likely and less likely outputs.

Interesting insight: By training language models to generate less likely hypotheses, decision-makers can overcome biases and consider a broader range of possibilities.

Read Article

Wrapping Up

Your engagement and insights continue to be the driving force behind AI Unchained's journey.

I'm always eager to hear your thoughts, so please don't hesitate to reach out by replying to this email or connecting with me on Twitter.

Until next week,

Joris