Chat GPT - Code Interpreter - Tips / Tricks / Documentation

ChatGPT
1

Firstly I want to say a massive well done to openAI, the new Code Interpreter is incredible across multiple complex domains. However there doesn’t seem to be much documentation on the module.

Here’s my top tips/tricks

  1. You have to enable it. under your profile in chatGPT, click beta and turn it on
  2. It will allow multiple files if you zip them first then upload. You can also add as you go along to compare or add.
  3. If you want an output, then make sure to explicitly state what output you want with your prompt, i.e. “export as a csv”, “generate a summary presentation for the executives”, “export as excel file”
  4. Take a copy of the steps, you can use this to run later
  5. If you are hitting memory warnings, send in a representative sample of your data. Use this to generate the steps and desired outcome. Use tip 4 to run on the big local dataset.

Please share your tips and ideas on use!

Cheers OpenAI! :clap:

6 Likes
3

This right here is a :star2: tip.

Well done!

1 Like
4

Do you feel its better than vanila gpt-4? I didn’t think so.

1 Like
5

Hi Thiago,

It seems to “think” better in our experiments, but milage may vary for your use case.

1 Like
6

Thanks for sharing! I am so excited to try this feature!

1 Like
7

You can use code interpreter to generate a mock dataset, generate a report and write it to Word doc and PowerPoint presentation check out this prompt that does all of the above:

Prompt:
Create a 2000 row dataset of employees. Use the columns a typical HR department will need and write them to a CSV file, then create a Excel spreadsheet. Create a tab called employees salaries in the Excel spreadsheet and populate it with data of the employee names and salaries.
Analyze, give insight and visualize the dataset. Create a well-formated report and write to Word doc and then create a PowerPoint presentation on the analyzed, visualized dataset with insights. Also, give me ten questions to ask this dataset.

Prompt 2:
Create a 2000 row dataset of a typical webserver log. Use the columns typical to a webserver, write to a csv file. Analyze and visualize the dataset. Also give me 10 questions to ask this dataset

3 Likes
8

hey! any recommendations on the prompt to get it to do this? My previous prompts are no longer working.

1 Like
9

If you mean access the code interpreter, it’s a beta feature available to plus subscribers that has to be enabled first.

1 Like