Lex Fridman Podcast Cristos Goodrow: YouTube Algorithm
22 snips
Jan 25, 2020 
Cristos Goodrow, VP of Engineering at Google and head of YouTube's algorithm, shares insights into the mechanics of content discovery on the platform. He discusses the balance between viewer engagement and moderation, revealing how user interactions shape personalized recommendations. Goodrow also dives into the challenges of clickbait and its effects on creators, as well as the complexities of creating viral content. He emphasizes the need for a healthy online community and explores YouTube's future as it transitions away from traditional TV.
AI Snips
Chapters
Transcript
Episode notes
Content Responsibility
- Establish clear content guidelines and boundaries on YouTube.
- Reduce recommendations of borderline content and promote authoritative sources.
Collaborative Filtering
- Collaborative filtering maps videos watched together, creating a "related graph".
- This helps group similar videos, even across different languages, without explicit tagging.
Bilingual Recommendations
- A researcher was surprised how YouTube could correctly recommend videos in different languages.
- Collaborative filtering naturally groups content by user behavior, even for bilingual users.