Episode
Cristos Goodrow: YouTube Algorithm
1:31:18
Published: Sat Jan 25 2020
Description
Cristos Goodrow is VP of Engineering at Google and head of Search and Discovery at YouTube (aka YouTube Algorithm). This conversation is part of the Artificial Intelligence podcast. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter, LinkedIn, Facebook, Medium, or YouTube where you can watch the video versions of these conversations. If you enjoy the podcast, please rate it 5 stars on Apple Podcasts, follow on Spotify, or support it on Patreon. This episode is presented by Cash App. Download it (App Store, Google Play), use code "LexPodcast". Here's the outline of the episode. On some podcast players you should be able to click the timestamp to jump to that time. 00:00 - Introduction 03:26 - Life-long trajectory through YouTube 07:30 - Discovering new ideas on YouTube 13:33 - Managing healthy conversation 23:02 - YouTube Algorithm 38:00 - Analyzing the content of video itself 44:38 - Clickbait thumbnails and titles 47:50 - Feeling like I'm helping the YouTube algorithm get smarter 50:14 - Personalization 51:44 - What does success look like for the algorithm? 54:32 - Effect of YouTube on society 57:24 - Creators 59:33 - Burnout 1:03:27 - YouTube algorithm: heuristics, machine learning, human behavior 1:08:36 - How to make a viral video? 1:10:27 - Veritasium: Why Are 96,000,000 Black Balls on This Reservoir? 1:13:20 - Making clips from long-form podcasts 1:18:07 - Moment-by-moment signal of viewer interest 1:20:04 - Why is video understanding such a difficult AI problem? 1:21:54 - Self-supervised learning on video 1:25:44 - What does YouTube look like 10, 20, 30 years from now?
Chapters
YouTube's recommendation system is one of the biggest in the world and has become a source of education and entertainment for millions.
00:00 - 02:52 (02:52)
Summary
YouTube's recommendation system is one of the biggest in the world and has become a source of education and entertainment for millions. This system is set to become one of the most impactful areas of AI in the 21st century.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The use of clustering and embedding techniques on YouTube video space can help maximize a person's happiness or growth as a human being by identifying trajectories that correlate with their interests and behaviors, according to experts.
02:52 - 11:06 (08:13)
Summary
The use of clustering and embedding techniques on YouTube video space can help maximize a person's happiness or growth as a human being by identifying trajectories that correlate with their interests and behaviors, according to experts.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The guest discusses about the challenges of curating speech and creating a better society without limiting freedom of speech too much, and how YouTube can reduce the meanness of its users.
11:06 - 23:23 (12:16)
Summary
The guest discusses about the challenges of curating speech and creating a better society without limiting freedom of speech too much, and how YouTube can reduce the meanness of its users.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The related graph on YouTube represents videos that are similar or related, with videos frequently watched by the same people placed closer together, and those that do not frequently get watched by the same people placed farther apart, making it easier for the platform to recommend videos to bilingual users.
23:23 - 30:42 (07:19)
Summary
The related graph on YouTube represents videos that are similar or related, with videos frequently watched by the same people placed closer together, and those that do not frequently get watched by the same people placed farther apart, making it easier for the platform to recommend videos to bilingual users.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
YouTube uses vectors in the space of all the videos on the platform to generate diverse recommendations for users based on video views, time watched, likes, comments, and more.
30:43 - 38:02 (07:19)
Summary
YouTube uses vectors in the space of all the videos on the platform to generate diverse recommendations for users based on video views, time watched, likes, comments, and more.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The video-making strategy that seems to work best is to create high-quality videos with click-baity titles and thumbnails to attract viewers.
38:02 - 48:14 (10:11)
Summary
The video-making strategy that seems to work best is to create high-quality videos with click-baity titles and thumbnails to attract viewers. However, even if you take the algorithm out of the picture, the titles and thumbnails of the videos still play a significant role in attracting viewers.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The speaker discusses the importance of personalization on YouTube and how liking a video can improve the user's experience and help the machine learning algorithm.
48:14 - 56:15 (08:01)
Summary
The speaker discusses the importance of personalization on YouTube and how liking a video can improve the user's experience and help the machine learning algorithm.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The YouTube algorithm is a combination of code, machine learning systems, and people's behavior on the platform.
56:15 - 1:05:56 (09:40)
Summary
The YouTube algorithm is a combination of code, machine learning systems, and people's behavior on the platform. Creators often experience burnout due to the highs and lows of social media, but it's important to take breaks and refresh oneself when going through psychological challenges.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
The success of YouTube's recommendation experiment can be determined by the decrease in video dismissals and an increase in viewer satisfaction and five-star ratings.
1:05:56 - 1:10:54 (04:58)
Summary
The success of YouTube's recommendation experiment can be determined by the decrease in video dismissals and an increase in viewer satisfaction and five-star ratings. The first diversity measure was limiting the number of videos from the same channel in a row, but viral videos now play a part in improving recommendations.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
In this podcast, a discussion ensues regarding the possibility of an algorithm that would be able to pinpoint exciting moments in a video to clip it out automatically.
1:10:54 - 1:18:08 (07:13)
Summary
In this podcast, a discussion ensues regarding the possibility of an algorithm that would be able to pinpoint exciting moments in a video to clip it out automatically. The implications of such technology and the current problems of algorithmic recommendations are examined.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
Discussing the possibility of an algorithm that watches numerous YouTube videos to predict the next frame of video, and whether it can understand enough about the physics of the world and perform common sense reasoning required for the challenge.
1:18:07 - 1:25:59 (07:51)
Summary
Discussing the possibility of an algorithm that watches numerous YouTube videos to predict the next frame of video, and whether it can understand enough about the physics of the world and perform common sense reasoning required for the challenge.
EpisodeCristos Goodrow: YouTube Algorithm
PodcastLex Fridman Podcast
While the future of YouTube may include virtual reality, the main goal is to continue to introduce people to the vast array of content available and to ensure that the platform is acting responsibly and enriching people's lives.
1:26:01 - 1:30:55 (04:54)
Summary
While the future of YouTube may include virtual reality, the main goal is to continue to introduce people to the vast array of content available and to ensure that the platform is acting responsibly and enriching people's lives.