S01 E05 - 005 - Bryan Cheong on Replika and RLHF

The KMO Show by KMO
Bryan Cheong, a machine learning engineer, joins KMO for a mostly non-technical discussion of large language models, reinforcement learning from human feedback, the democratizing potential of the current moment in artificial intelligence, and why your shouldn't get attached to a centrally-controlled AI companion when you can make and run  ...  See more
Mar 30 2023

Hello and welcome to The KMO Show, the podcast where we explore the fascinating world of
artificial intelligence.
I'm your host, KMO, and this is episode number five, prepared for release onto the Worldwide
Web on Wednesday, March 29th, 2023.
Today I have a very special guest with me.
He is Brian Chong, a machine learning engineer in San Francisco.
His previous work is in ML for forecasting and materials optimization.
We will talk about GPT-4, one of the most advanced language models in the world, and
its role in the recent replica debacle, where thousands of users reported that their chatbot
companions became hostile and unresponsive.
But before we get to that, let me explain what a language model is and how it works.
A language model is a computer program that can generate text based on

See full transcription
reinforcement learning