2min chapter

Machine Learning Street Talk (MLST) cover image

Neel Nanda - Mechanistic Interpretability

Machine Learning Street Talk (MLST)

CHAPTER

What Is Mechanistic Interpretability?

Neil has an amazing YouTube channel I've been glued to it all week actually some of them are admittedly quite technical but even if you're not interested in mechanistic interpretability Neil has an extremely soothing voice second only to Sam Harris. By the way there was something I didn't say in the woods which is that Neil's dulce it tones will melt the stress away quicker than a nun's first curry. Okay so we're now going to kick off with the actual main podcast so that was a nice little piece but maybe we'll put that on the patreon's thing magic. Anyway with that said we started to talk about what is mechanistic interpretable and first of all I wanted to

00:00
Speaker 2
Okay so we're now going to kick off with the actual main podcast so that was a nice little piece but maybe we'll put that on the patreon's thing magic. By the way there was something I didn't say in the woods which is that Neil has an amazing YouTube channel I've been glued to it all week actually some of them are admittedly quite technical but even if you're not interested in mechanistic interpretability Neil has an extremely soothing voice second only to Sam Harris and I would recommend listening to him when you go to sleep because as you know Neil's dulce it tones will melt the stress away quicker than a nun's first curry. Anyway with that said we started to talk about what is mechanistic interpretability and first of all I wanted to call out your ridiculously detailed and exquisite mechanistic interpretability explainer maybe you could just tell us about that quicker.
Speaker 1
Yes so I wanted to try to write a glossary there's some basic common terms and mechanism there's like an appendix to a blog post. There were a lot of terms and mechanism there are a lot of terms and mechanism and I like writing it and very bad at privacy so I got kind of carried away and there's now a 33,000 words massive massive exposition but importantly it is designed to be easily such and mechanistic is full of jargon and I'm sure I'll forget to explain everything that I'm saying so I'd highly recommend just having it open an attack as you listen to this and if you get lost just look up terms in there and yes it's both definitions but it's also long tangents giving intuitions and context and related work and con misunderstandings it was very fun to write.

Get the Snipd
podcast app

Unlock the knowledge in podcasts with the podcast player of the future.
App store bannerPlay store banner

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

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

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

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

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

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

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode