type
Post
Created date
Apr 16, 2023 11:04 AM
category
Data Science
tags
Machine Learning
Programming
Artificial Intelligence
status
Published
Language
English
From
Seminar
summary
At a recent event on Large Language Models (LLMs), Sam Witteveen, a Google Developer Expert for Machine Learning, spoke about the latest advancements in LLMs, including ChatGPT, GPT-4, and Bard, and other semi-open source models with different training regimes and datasets.
slug
Google-seminar-llm
password
Author
Priority
Urgent + Important
Featured
Featured
Cover
Origin
Type
Youtube
Youtube
icon
I joined a event on 13/04. At the event about Large Language Models (LLMs), Sam Witteveen, a Google Developer Expert for Machine Learning, spoke about the latest advancements in LLMs, including ChatGPT, GPT-4, and Bard, and other semi-open source models with different training regimes and datasets.
notion image
He highlighted the best models and techniques that set them apart from their predecessors, including the "Transformer" architecture, and "RLHF" or "Reinforcement Learning from Human Feedback". The event also showcased companies and organizations involved in LLM development, such as OpenAI, Google Chinchilla, Meta, Databrick, Cerebras, ElutherAI, Raven, Vicuna, and Koala.
In addition to discussing models and techniques, Sam also shared best practices for building LLM apps and APIs using frameworks like Langchain and Llama Index. He emphasized the value of using open source tools such as "Flan 20B", "LangChain", and "AutoGPT" for fine-tuning LLMs for specific tasks. Overall, the event provided a valuable learning experience for those interested in AI and natural language processing.
 

Here are key slides I took on the event:

Resources shared by Sam

 
How Ali Abdaal Manage My Time - The Trident Calendar System尋找人生目標: A goal without a plan is just a wish