Data Science and Coding with LLMs

This workshop, "Data Science and Coding with LLMs," is designed for college students with some prior experience in data science and coding. Here's a summary of the key topics and takeaways:

● The workshop will focus on practical coding examples that are up-to-date, using tools like GPT-4 and Gemini. These examples are real-life scenarios, generated since Fall 2024.

● Participants will learn to utilize ChatGPT/Gemini for specific coding tasks, rather than trying to build complete applications. The workshop will cover how to use prompt and choose the right coding libraries for specific purposes, which can increase productivity. You will learn to refine questions through conversation to get closer to the desired output.

● Emphasis will be placed on code reliability. The workshop will stress that it is crucial to verify code generated by AI and not trust it 100%.

● Debugging techniques will be covered, including checking error logs, isolating the bug, and making assumptions about the errors to test them. 

● The workshop will emphasize the importance of well-documented and readable code for portability, usability, and maintainability. It will also be noted that programmers spend time understanding, reusing, and modifying existing code, and that hard-to-read code leads to bugs and inefficiencies.

● Attention will be given to the importance of indentation in coding, particularly in Python, where it is critical to the structure of the code. Copy/pasting Python code can lead to problems.

● The workshop will touch on the issue of hallucination in LLMs.

● Other topics include: file I/O operations in Python and R, writing small scripts such as shell scripts and Google Apps Script.

The workshop will stress that asking the right questions is fundamental. It will also underscore that most of today's programming consists of re-using existing code in slightly different forms, and that high levels of reasoning may not always be required.