Part 4/11:
Shapiro demonstrates how ChatGPT can assist in scripting these transformation steps. He prompts GPT-3 to generate code that opens JSON files, validates them, handles exceptions gracefully, and converts valid data into GEXF format. His iterative process involves debugging, adjusting code, and making the scripts more robust—adding features like consistent UTF-8 encoding and error handling.
A crucial insight emerges: ChatGPT expedites the generation of code, but it can also hallucinate or invent modules that don’t exist. For instance, it suggested validation functions that aren’t real, illustrating the importance of a developer’s oversight. This highlights a key theme—AI as an enabler that speeds up coding but still requires human judgment.