ChatGPT Goes Beyond Its Knowledge Cut-Off
summary
Live-coding that scrapes TechCrunch articles from the past seven days, extracts paragraphs with BeautifulSoup and pandas, and reduces ~2,300 rows to a smaller set by dropping paragraphs under ~10 words; paragraphs get OpenAI embeddings (via the OpenAI/Bing client shown) stored in a vector DB, queries use vector similarity (cosine) to retrieve the top 1–3 paragraphs, and those passages are prepended to a GPT prompt so the model answers with up-to-date TechCrunch content; live demo shows the pipeline running end-to-end and handling edge cases like noisy ads and pagination with a Jupyter notebook.
AI-written summary · about
chapters (9)
- 0:00Introduction and setup
- 2:09Introducing RAG concept
- 3:57Fetching news articles
- 8:19Processing article content
- 13:49Cleaning and structuring data
- 16:45Vector embedding with OpenAI
- 20:40Querying with embeddings
- 22:30Answering the user question
- 27:01Conclusion and Q&A
related talks
01h 09mChatGPT and OpenAI API solutions: 👑 My Favourite Implementations
45mHow to use GPT in your projects: A Case Study
01h 00mGoing beyond ChatGPT: introduction to prompt engineering & LLMs
01h 16mChatGPT - and How we Can Benefit from it
02h 00mChatGPT & Prompt Engineering
01h 12mChatGPT and OpenAI API solutions: 👑 My Favourite Implementations
27mChatGPT Ignites
28mGPT or how I learned to stop worrying and love the generative AI