🧠 Code Interpreter is a treasure chest

Here's how to maximize it

The TLDR

Main Project: After testing out Code Interpreter (ChatGPT) for a month, we’re hooked on using it for data analysis and supercharged pattern recognition. This week, we’re sharing quick tips on how to maximize it and how you can start incorporating it today.

Musing of the Week: Code Interpreter represents a huge step forward in closing skill gaps and making knowledge work accessible. We consider what similar tools mean for the creation of new possibilities and potential.

Data Analysis and Visualization Created from CSV using Code Interpreter

 ðŸ’¬ The Power Of Code in Conversation

What is it?

Code Interpreter is a plugin (by OpenAI) that augments ChatGPT’s capabilities. It’s available to all ChatGPT Plus users ($20/month), and rolled out in beta for all Plus users about in July 2023. Now that we’ve had time to test it out for ourselves, we’re sharing some tips for maximizing the tool.

The Value Prop:

Code Interpreter acts as a really great technical analyst. The tool allows you to upload files and execute code within ChatGPT, meaning you don’t need robust technical knowledge to take advantage of data analysis or get the most out of the information you collect. All you need is some data/documents and a clear idea of what you’re trying to accomplish, give ChatGPT access to your documents, and it will help you accomplish your analysis just by conversing.

Ensuring Good Output from Code Interpreter:

  • The biggest limitation is that it doesn’t retain data beyond the current conversation. You’ll need to give ChatGPT some context each time you start a conversation, so create an accessible file where you store your set-up prompt and documents. Upload this at the start of each relevant session.

  • Ask ChatGPT to export anything you’ve learned at the end of a session so that you can re-upload it next time to pick up where you left off.

  • Don’t stop ChatGPT in the middle of Code Interpreter responses. This almost always causes an error on your next prompt that requires a page reload.

Common Use Cases to Test - Try it out!

  • Data Cleaning: Ask it to identify missing or duplicate data

  • Data Analysis: Upload 2 datasets and ask for correlations/patterns

  • Data Visualization: Ask for specific types of tables and graphs

  • QR Code Creation: Ask for a QR code for a specific webpage

  • Template Customization: Pass it a template document and a context document, and ask it to fill in the blanks to customize your template

  • Even more cool use cases :) linked here 

Example - Comparing Data Sets

You can see our experiment comparing two data sets below. We randomly chose two data sets about global cities (temperature and cost of living) to see if they had any correlation. The whole process took less than five minutes, from discovering the data to having fully realized visualizations with interpretation from ChatGPT. (Data from Kaggle)

Visualizing the Correlation Between Average Temperature and Cost of Living

🧠 Musing of the Week

Code Interpreter excites us because it bridges the gap between different sides of our workflow. Instead of shifting our momentum and setting up a development environment to grind through a backlog of data, we can simply have a conversation with ChatGPT and walk away with a robust set of insights. This makes it less daunting to build complex analyses and leads to a more robust work product.

Moreover, it speaks to AI’s broader ability to expand human potential. As AI continues to advance, we hope to see more integration of technical tools in fields that have been traditionally more purely qualitative and improved communication in technical fields that sometimes struggle with conveying ideas to broader audiences. There will always be space and a need for specialization, but lowering the threshold to have accessible high-yield tools like Code Interpreter will change the landscape and allow for new intersections and collaborations between disciplines.

🙌 If you’re hyped about the generative AI industry specifically, here are some of the coolest roles we’ve seen this week:

🔨 Check out these other AI tools we’ve been playing with this week:

  • Musico: Limitless copyright-free AI generative music

  • Review.io: Quickly craft thoughtful responses to customer reviews

  • Microsoft Designer: Generate beautiful designs (e.g., social media graphics) with one prompt

That’s all for this week. See you next Tuesday!

Lorel & Reily