Last, I set out to use ChatGPT to generate content about NBA games. However, I only used static strings with dummy data, and this process will not scale to the level of automation that will shape the future of journalism. The next evolution of this project would be to test these frontiers and would include dynamically passing data from a cloud-based data table.

Here’s my mission statement:

Within the next month, I will combine a Python based data pipeline and the generative capability of ChatGPT. Signs of my progress will include being able to create game information in Python code, being able to generate new ChatGPT queries with just a game identifier provided, and being able to get relatively standardized output as per instructions.

Here’s my action statement:

The activities I will conduct are identifying the dynamic inputs, generating the data in the cloud, transforming the data to generate facts for a specific game, and testing the output from ChatGPT.

Project Plan:

• Week 1
  • Write monthly introduction post
  • Connect to cloud-based data tables in a Python environment
• Week 2
  • Identify all dynamic string inputs
• Week 3
  • Write a Python script to fill in as many of the dynamic fields as possible
  • Test script to make sure it produces dynamic string desired, which can be passed to ChatGPT API
• Week 4
  • Produce article for one game, generated by a Python script which uses data in a live cloud-based data source
  • Write M2M recap post

I’ll check in at the end of the month. Happy August.