Learn how to enhance GPT-4's capabilities and overcome its limitations through retrieval augmentation in this 27-minute video tutorial. Explore techniques to address hallucinations and outdated information in Large Language Models by combining OpenAI's ChatCompletion endpoint with the Pinecone vector database. Follow along as the process of augmenting GPT-4's knowledge is demonstrated using the LangChain Python library as an example. Discover methods for scraping documentation, preprocessing text, creating embeddings, and performing semantic searches to retrieve relevant information. Compare the performance of GPT-4 with and without augmentation, and gain insights into building powerful AI tools with ease.
Overview
Syllabus
Why GPT-4 can fail - hallucinations
What we can do with retrieval augmentation
How retrieval augmentation works
Scraping docs for LLMs
Preprocessing and chunking text for GPT4
Creating embeddings with text-embedding-ada-002
Creating the Pinecone vector database
Retrieving relevant docs with semantic search
GPT-4 generated answers
GPT-4 with augmentation vs. GPT-4 without
Building powerful tools is almost too easy
Taught by
James Briggs
Related Courses
-
Fixing LLM Hallucinations with Retrieval Augmentation in LangChain
-
OpenAI's New GPT 3.5 Embedding Model for Semantic Search
-
Augmented Language Models - LLM Bootcamp
-
GPT-4 & LangChain Tutorial - How to Chat With A 56-Page PDF Document
-
Better Llama with Retrieval Augmented Generation - RAG
4.0 -
Beyond AI Vector Database - Neural Search Implementation with GPT-4 and ChatGPT
