💾 Archived View for bjornwestergard.com › notes › llm.gmi captured on 2023-07-22 at 16:28:08. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2023-05-24)
-=-=-=-=-=-=-
The best known examples of these at the time I write this are OpenAI's ChatGPT models, which can be accessed through an API at a cost proportional to the number of words in the prompt and response.
One interesting technique combines text generation and "embedding" (i.e. generating a vector representation of a document) to create a chatbot that can search a corpus of documents.
TypeChat (Typescript Types + LLM)
Simon Willison - Q&A against your documentation with GPT3, embeddings and Datasette
Langchain Retrieval Augmentation ("Fixing Hallucination with Knowledge Bases")
Random Projection for Locality Sensitive Hashing
Writing an LLM Agent in 50 Lines of Code
Martin Fowler: Example of LLM prompting for programming
Ed Frank - Rebuttal to Society’s Tech Debt & Software’s Gutenberg Moment