Transcription APIs and Integrating LLMs with Monarch Wadia
Anthony Campolo discusses his work on integrating transcription APIs into a workflow and explores options for managing chat histories. Monarch Wadia introduces his library Ragged, which aims to simplify and standardize the process of working with large language models (LLMs) in JavaScript and TypeScript projects.
00:00 - Introduction and Overview of Episode Topics
Anthony and Monarch introduce the main topics they will be discussing in the episode, including transcription APIs, managing chat histories, and Monarch's ragged library for working with LLMs.
02:56 - Monarch's Job Search and the AI Job Market
Monarch shares his experiences applying for jobs at AI companies and discusses the current state of the job market in the AI industry.
09:25 - The Sydney Chatbot and Prompting Personalities
Anthony and Monarch discuss the Sydney chatbot incident with Bing and explore the idea of prompting AI to take on different personalities or archetypes.
20:36 - Anthony's Work on Transcription APIs
Anthony shares his progress on integrating various transcription APIs into a workflow, comparing features, pricing, and ease of use across different providers such as Assembly AI, DeepGram, and Speechmatics.
30:29 - Web LLM and Running Language Models in the Browser
Monarch introduces Web LLM, a library that allows running large language models directly in the browser, and discusses the implications and potential use cases for this technology.
37:39 - Monarch's ragged Library and Simplifying LLM Integration
Monarch explains the motivation behind creating ragged, a library that aims to simplify and standardize the process of working with LLMs in JavaScript and TypeScript projects, and how it differs from existing solutions like Langchain.
57:03 - The use-ragged React Hook and Managing Chat Histories
Monarch demonstrates how to use the use-ragged React hook provided by his library and explains how it manages chat histories and simplifies the process of working with LLMs in a React application.
1:11:37 - Comparing ragged to Other LLM Libraries
Anthony and Monarch explore other libraries for working with LLMs, such as LlamaIndex and LLM Client, and discuss how they compare to ragged in terms of features, API design, and ease of use.
1:19:00 - Tangents on Science, Hermeticism, and the Development of Human Consciousness
The conversation veers into various tangents, touching on topics such as the scientific method, the symbolism of the caduceus in Hermeticism, and the role of snake venom in the development of human consciousness.
1:27:31 - Wrapping Up and Plans for Next Week
Anthony and Monarch summarize their progress and set goals for what they plan to work on and discuss in the next episode.
Links and Resources
https://www.nytimes.com/2023/02/16/te...
https://en.wikipedia.org/wiki/Multipl...
/ web-llm-lets-you-run-llms-natively-in-your...
/ ghost-in-the-machine-a-glimpse-into-gpt-4-...
https://en.wikipedia.org/wiki/Jungian...
https://www.assemblyai.com/
https://deepgram.com/
https://www.speechmatics.com/
https://ollama.com/
https://webllm.mlc.ai/
https://github.com/jzhang38/TinyLlama
https://secretllama.com/
https://fal.ai/
https://rxdb.info/
https://github.com/monarchwadia/ragge...
https://python.langchain.com/docs/use...
https://python.langchain.com/docs/exp...
https://docs.llamaindex.ai/en/stable/...
https://ts.llamaindex.ai/api/classes/...
https://github.com/dosco/llm-client
https://arxiv.org/pdf/2201.11903
https://www.vectorsofmind.com/p/the-s...