We discuss: AI Coding Tools, Code-Interpreter, Python, Regex, Network errors, Machine Learning, ChatGPT4, HTML Parsing, Github Copilot vs GPT-4
In this episode, we delve into a fascinating experiment where I, Alex Denne, under the watchful guidance of ML Research Scientist Alex Pap, try to get AI to writing some regex that can be run locally on my machine using python, on millions of documents.
The goal? To extract matching text from millions of HTML files.
It all inadvertently unfolds into an intriguing journey of trial and error.
For the no-code listeners, this episode offers first-hand insights into the application and limitations of AI coding tools and code interpreters (and why, for now, you probably still need technical help like Alex D did!)
At the outset, we were greeted by a seemingly promising result - a neat CSV file with the right column names but no entries as the AI successfully claimed to extract definitions only to produce an empty result.
In an attempt to further probe, the AI was prompted to read the first 100 characters for potential matches. Alas! In lieu of any found matches, it concluded the document must be lengthy and gracefully tapped out.
In addition, we had to deal with several network errors that may be attributed to the reported DDoS attacks on OpenAI.
After multiple hits and misses, we decided to start afresh with a new approach. We didn't exactly strike gold, but we learned a lot.
Through this episode, we touch upon topics like ChatGPT4 and the wonderful feature of 'dragging and dropping' files into GPT-4 Turbo.
Watch USING AI on youtube: