Parabolica #15 - How I combined ChatGPT and my own blog posts

Time is the ultimate currency.

Stop wasting your life away looking for that document containing that one statistic you vaguely remember from 8 months ago.

“Is our y1-y2 retention rate 34% or 45%?”

Just ask the AI, silly.

I want to show you how to build an AI chatbot, powered by ChatGPT, that uses your own data.

A few weeks ago, I wrote about AI chatbots built on your own data in this post.

I decided to follow the Google Colab instructions from Dan Shipper who created chatbots for Lenny’s Newsletter and Andrew Huberman.

It was a little out of date so I made some updates to the code.

I figure many of the companies offering to create chatbots are powered by OpenAI.

It’s also important to understand how the underlying technology works!

---

So here are the basic steps to create your own chatbot with OpenAI:

1. Download the text you want to embed the AI with

2. Install OpenAI and LangChain dependencies

3. Define functions you will use later

   1. construct_index()

   2. ask_chatbot()

4. Set OpenAI API Key

5. Construct_index() - costs $

   1. This is where you pay OpenAI to embed your text or “index”

6. Ask away with ask_chatbot() - costs $

---

Chatbot in action

Next steps to improve the chatbot:

1. create a pretty webpage for it

2. use pinecone, a vector database, to store the embeddings

---

Takeaways for Marketing

You can do this with your own knowledge base and chat with it.

Here are some sample questions:

• What’s the best creative concept last quarter?

• What’s our best performing headline?

• What’s a good trial to subscribe rate?

• What’s a good install to trial rate?

• What’s the average CPI in the US?

• Who’s our highest LTV demographic?

• What are our best performing days of the week/hours of the day?

How many hours a month will a customized AI chatbot save your org?

Stop wasting time. Combine the power of ChatGPT and your own data.

Ping me at manson@goparabolica.com for help.