AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
3min chapter
#126 - few have tried ChatGPT, Baidu’s $145M AI fund, AI job culling, OpenAI grant program
Last Week in AI
Reward Collapse in Aligning Large Language Models
There are two different kinds of prompts that we give to our AI systems. One is an open-ended question like what is the capital of Canada right now and then there's a clear answer to that second category. These systems these RLHF schemes they usually rely on having humans rank a bunch of different responses that a language model might give to a question like that. Some of those answers are going to be completely right and should get full reward and some of them are just going to be laughably wrong but it turns out that when you fail to account for that subtle difference your system ends up treating all of these inputs in the same way which leads to less effective models.
00:00
Transcript
Episode notes
Get the Snipd
podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered
Discover
highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any
moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share
& Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Discover
highlights
Listen to the best highlights from the podcasts you love and dive into the full episode