
41 segments available
Cristos Goodrow is VP of Engineering at Google and head of Search and Discovery at YouTube (aka YouTube Algorithm). This episode is presented by Cash App. Download it & use code "LexPodcast": Cash App (App Store): https://apple.co/2sPrUHe Cash App (Google Play): https://bit.ly/2MlvP5w PODCAST INFO: Podcast website: https://lexfridman.com/podcast Apple Podcasts: https://apple.co/2lwqZIr Spotify: https://spoti.fi/2nEwCF8 RSS: https://lexfridman.com/feed/podcast/ Full episodes playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOdP_8GztsuKi9nrraNbKKp4 Clips playlist: https://www.youtube.com/playlist?list=PLrAXtmErZgOeciFP3CBCIEElOJeitOr41 OUTLINE: 0:00 - Introduction 3:26 - Life-long trajectory through YouTube 7:30 - 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? CONNECT: - Subscribe to this YouTube channel - Twitter: https://twitter.com/lexfridman - LinkedIn: https://www.linkedin.com/in/lexfridman - Facebook: https://www.facebook.com/LexFridmanPage - Instagram: https://www.instagram.com/lexfridman - Medium: https://medium.com/@lexfridman - Support on Patreon: https://www.patreon.com/lexfridman
Cristos Goodrow discusses the immense scale of YouTube, highlighting its role as the second most popular search engine and a platform for learning. He reflects on the transformative impact of YouTube on education and personal growth, emphasizing the responsibility of users in their online journeys. Goodrow also introduces the challenges faced by the YouTube algorithm in curating content effectively.
"the following is a conversation with Christos Kudrow vice president of engineering at Google and head of search and discovery at YouTube also known as the YouTube algorithm YouTube has approximately 1..."
In this segment, Goodrow explores the potential trajectories through YouTube that can enhance happiness and education. He shares personal anecdotes about how his children have benefited from YouTube, illustrating the platform's positive influence on learning and character development. The discussion emphasizes the importance of evolving interests and the need for YouTube to adapt to users' changing preferences.
"most popular search engine behind Google of course we watch more than 1 billion hours of YouTube videos a day more than Netflix and facebook video combined YouTube creators upload over 500 thousand ho..."
Goodrow delves into the complexities of introducing diversity in YouTube's content recommendations. He explains the challenge of balancing user interests with the need for exposure to new ideas. By clustering videos and analyzing user behavior, YouTube aims to recommend content that broadens viewers' horizons while maintaining engagement.
"continue to enrich people's lives then you know then it has to grow with them and and people's interests change over time and so I think we've we've been working on this problem and I'll just say it b..."
This segment addresses the challenges of managing political content on YouTube. Goodrow discusses the responsibility of YouTube to ensure that diverse viewpoints are represented while also maintaining standards for credible information. He highlights the importance of authoritative sources in fostering healthy discourse and navigating the complexities of political ideologies.
"talk about the machine learning of that but I have to linger on things that neither you or anyone have an answer to there's gray areas of truth which is for example now I can't believe I'm going there..."
Goodrow tackles the issue of negativity and trolling on YouTube. He acknowledges the need for creators to develop resilience against online criticism and discusses YouTube's efforts to reduce meanness through comment ranking and user controls. The segment emphasizes the ongoing challenge of fostering a positive community while allowing freedom of expression.
"settle that choose a side or anything like that what we're trying to do is make sure that the the people who are expressing those point of view and and offering those positions are authoritative and c..."
In this concluding segment, Goodrow reflects on the interplay between machine learning algorithms and human oversight in curating content on YouTube. He emphasizes the necessity of both approaches to effectively address misinformation and uphold community standards. The discussion highlights the critical role of human judgment in shaping the future of content moderation.
"comments based on whatever based on how much they contribute to the healthy conversation let's put it that way then the other is almost an interface question of how do you how does the Creator filter ..."
In this segment, Goodrow explores the interplay between machine learning algorithms and human intervention in content curation on YouTube. He explains how human decisions inform the algorithms, particularly in identifying misinformation and policy violations, highlighting the necessity of both elements in the content moderation process.
"people here YouTube in a room sitting and thinking about what is the nature of truth what is what are the ideals that we should be promoting that kind of thing so algorithm versus human input what's y..."
Goodrow addresses the potential biases in human content reviewers on YouTube. He explains the measures taken to mitigate these biases, such as promoting scientific consensus and ensuring diverse backgrounds among reviewers, while acknowledging the challenges of bias in machine learning systems.
"you have a sense that these human beings have a bias in some kind of direction sort of I mean that's the interesting question we do sort of in autonomous vehicles and computer vision in general a lot ..."
Cristos Goodrow introduces the basics of the YouTube algorithm, explaining how it recommends videos based on user behavior and search queries. He discusses the evolution of the algorithm and its ability to improve recommendations over time, emphasizing the importance of user engagement metrics.
"protected class for instance thank you for exploring with me some of the more challenging things I'm sure there's a few more that we'll jump back to but let me jump into the fun part which is maybe th..."
In this segment, Goodrow explains collaborative filtering, a technique used by YouTube to recommend videos. He describes how the algorithm analyzes viewing patterns to create a related graph of videos, allowing for more accurate recommendations based on user behavior.
"but there are much more sophisticated things where we're mostly trying to do some syntactic match or or maybe a semantic match based on words that we can add to the document itself for instance you kn..."
Goodrow discusses the significance of user history in shaping YouTube recommendations. He shares insights on how the platform clusters videos based on user preferences and behaviors, illustrating the potential for personalized content discovery through understanding individual viewing patterns.
"well I'm a researcher in in the US and and when I'm looking for academic topics I want to look I want to see them in English and so she searched for one found a video and then looked at the watch next..."
In this segment, Goodrow delves into how YouTube measures video quality and user satisfaction. He explains the transition from simple view counts to more nuanced metrics, including watch time and user feedback, to better assess the value of content and improve recommendations.
"on YouTube then you can start to say okay well you know which videos which which other vectors are close to me and to my vector and and that's one of the ways that we generate some diverse recommendat..."
Goodrow shares insights into user subscription behaviors on YouTube, explaining how different users interpret the subscribe button. He discusses the complexities of why some users subscribe without engaging with content, and how this affects the algorithm's understanding of user preferences.
"something else well you mentioned commenting also sharing the video if you if you think it's worthy to be shared with someone else you know within YouTube or outside of YouTube as well either let's se..."
This segment focuses on the critical role of video titles and descriptions in YouTube's search and recommendation systems. Goodrow emphasizes that clear and relevant metadata is essential for both algorithmic visibility and user engagement, using examples to illustrate the impact of effective titling.
"then I like this person I really want to support them that that's how I click Subscribe right even though I may never actually want to click on their videos when they're releasing it I just love what ..."
Goodrow discusses YouTube's ongoing efforts to analyze video content itself, acknowledging the current limitations in accurately categorizing videos. He explains how this analysis can enhance searchability and improve recommendations, stressing the need for creators to provide clear titles and descriptions.
"the content analyzing the content while video right and what I can tell you is that our ability to do it well is still somewhat crude we can we can tell if it's a music video we can tell if it's a spo..."
In this segment, Goodrow addresses the tension between using clickbait titles and maintaining quality content. He compares different approaches to titling videos, discussing how creators can attract viewers while ensuring that the content delivers on its promises.
"away from wit and humor so you have to play with both right so but you're saying that for now sort of the content of the title the content of the description the actual text is is one of the best ways..."
Goodrow explores the concept of collaborative filtering on YouTube, explaining how user intent influences content discovery. He discusses the importance of user behavior in shaping recommendations and the challenges posed by attempts to manipulate the algorithm, emphasizing the need for genuine engagement.
"stuff discoverable I think is what you're really working on and hoping so yeah so from your perspective to put stuff in the description and remember the collaborative filtering part of the system it s..."
Goodrow outlines what success looks like for YouTube's algorithm, emphasizing user retention as a key indicator. He explains that if users return to watch more videos, it signifies that they find value in the content. He also discusses the importance of user satisfaction, as indicated by surveys, in determining the algorithm's effectiveness.
"life will be better and that's that kind of reasoning I'm not sure what that is and I'm not sure how many people share that feeling it could be just a machine learning feeling but at that point how mu..."
In this thought-provoking segment, Goodrow shares his vision of an ideal YouTube experience where every video watched is the best one ever. He reflects on a personal experience with a particularly impactful video, illustrating the emotional connection users have with content and the desire for continuous improvement in video recommendations.
"what a success look like in terms of the algorithm creating a great long-term experience for a user or put another way if you look at the videos I've watched this month how do you know the algorithm s..."
Goodrow discusses the broader societal implications of YouTube, highlighting its role in promoting openness and accessibility. He notes that unlike traditional media, YouTube allows anyone to upload content, which democratizes information and provides opportunities for creators from diverse backgrounds, especially in regions with low literacy.
"so on well so that's that's that's a heck of uh the thought that's one of the most beautiful and ambitious I think machine learning tasks so when you look at a society as opposed to any individual use..."
In this segment, Goodrow emphasizes the importance of the relationship between creators and their audiences on YouTube. He explains that users often express their love for the platform through their connection to specific creators or communities, rather than the technical aspects of the algorithm, underscoring the human element of content creation.
"society already there's a lot of what do you mean by openness well the fact that unlike other mediums there's not someone sitting at YouTube who decides before you can upload your video whether it's w..."
Goodrow reflects on the growth journey of YouTube creators, noting that many start with humble beginnings and evolve over time. He discusses how YouTube aims to support creators in building their audiences and fostering meaningful connections, which is essential for the platform's long-term success.
"changing society so I've worked at YouTube for eight almost nine years now and it's fun because I meet people and you know you tell them where they where you work you say you work on YouTube and they ..."
Goodrow tackles the issue of burnout among YouTube creators, acknowledging the psychological pressures they face. He reassures creators that taking breaks is not detrimental to their channels and can actually lead to improved content quality upon their return. This segment highlights the importance of mental health in the creative process.
"realize that YouTube is really about the video and connecting the people with the videos and then everything else kind of gets out of the way so beyond the video it's an interesting because you kind o..."
In this insightful discussion, Goodrow explains the complexity of the YouTube algorithm, which consists of various systems that interact with user behavior. He clarifies that the algorithm is not a single entity but a combination of code, machine learning, and human input, emphasizing the critical role of user engagement in shaping recommendations.
"creative ideas that someone has okay I think it's a really important thing to sort of to dispel I think it applies to all of social media like literally I've taken a break for a day every once in a wh..."
In this segment, Goodrow explains the significance of A/B testing in YouTube's decision-making process. He details how experiments are conducted to measure user satisfaction and engagement, using various metrics to determine the success of changes made to the platform.
"things in in sort of more fine-grained situations and I and I think that this is the way that the recommendation system and the search system and and probably many machine learning systems evolve is y..."
Goodrow addresses the difficulty of predicting which videos will go viral. He shares insights on past attempts to analyze view count trends and the factors that contribute to a video's viral potential, emphasizing the unpredictability of audience engagement.
"over time and so I think that just like with diversity you know I think the first diversity measure we took was okay not more than three videos in a row from the same Channel right it's a pretty simpl..."
This segment focuses on how YouTube analyzes viral videos after they gain popularity. Goodrow discusses the metrics used to understand viewer engagement and the factors that led to a video's success, including the sources of traffic and audience demographics.
"thing well you mentioned that a B experiments and so just about every single change we make to YouTube we do it only after we've run a a B experiment and so in those experiments which run from one wee..."
Goodrow explains the mechanics behind YouTube's recommendation system, detailing how it identifies and promotes content based on viewer preferences. He describes the iterative process of expanding recommendations to reach a broader audience while maintaining viewer satisfaction.
"it's improving the situation for viewers but we can also look at other things like we might do user studies where we invite some people in and ask them like what do you think about this what do you th..."
In this segment, Goodrow reflects on the elusive nature of creating viral content. He shares insights on the unpredictability of virality and the factors that can influence a video's success, including timing and audience engagement.
"oftentimes we look at where the traffic was coming from you know if it's if it's a lot of the viewership is coming from something like Twitter then then maybe it has a higher chance of becoming viral ..."
Goodrow discusses Derek Muller's viral video, 'Why Are 96,000,000 Black Balls on This Reservoir?' He analyzes the factors that contributed to its success, including viewer engagement and the video's unique appeal, illustrating how certain content resonates with audiences.
"like like I mentioned I hung out with Derek Muller a while ago a couple of months back he's actually the person who suggested I talk to you on this podcast all right well thank you Derek at that time ..."
This segment explores how YouTube's algorithm amplifies viral videos through recommendations. Goodrow explains the process of identifying and promoting content that aligns with viewer interests, creating a feedback loop that enhances engagement.
"like it I mean we can surely see where it was recommended where it was found who watched it and those sorts of things so it's actually sorry to interrupt it is the video which helped me discover who D..."
Goodrow shares his vision for the future of content discovery on YouTube, discussing the potential for automated clipping of interesting video segments. He emphasizes the importance of understanding video content to enhance user experience and facilitate easier access to engaging material.
"happened now you asked me about how to make a video go viral or make a viral video I don't think that if you or I decided to make a video about 96 million balls that it would also go viral it's possib..."
In this concluding segment, Goodrow addresses the challenges of deep video analysis and understanding viewer engagement. He reflects on the current limitations of algorithms in identifying compelling content and the potential for future advancements in this area.
"the the podcasts are doing yeah do you see as opposed to like I also add time stamps for the topics no people want the clip do you see YouTube somehow helping creators with that process or helping con..."
Cristos Goodrow shares his experiences with VR video on YouTube, discussing how heat maps reveal viewer attention and interest. He contrasts the engagement with a lecture video versus more dynamic content, highlighting the challenge of measuring viewer excitement and the potential for deeper insights into audience behavior.
"this a few times so I've uploaded myself it's a horrible idea some people enjoyed it but whatever the video of me giving a lecture in 360 over 360 camera it's cool because YouTube allows you to then w..."
Goodrow explores the complexities of determining what moments in videos captivate viewers. He reflects on personal experiences with impactful content, questioning how to effectively capture and analyze viewer reactions to identify exciting moments in videos.
"how you get that from people just watching except they tuned out at this point like it's hard to measure this moment was super exciting for people I don't know how you get that signal maybe comment is..."
The discussion shifts to the broader challenges of video understanding within the machine learning community. Goodrow emphasizes the intricacies of classifying video content and the difficulties in discerning significant moments, comparing it to natural language understanding.
"people to just label it yeah you mentioned that we're quite far away in terms of doing video analysis deep video analysis ago of course Google YouTube you know we're quite far away from solving autono..."
Goodrow introduces the concept of self-supervised learning as a potential solution for enhancing video understanding. He discusses the idea of predicting future frames in video content and how this approach could lead to a deeper comprehension of reality and common-sense reasoning.
"often pretty small right like you know you need to see this person's number in order to know which player it is and and there's a lot of players or you need to see you know the logo on their chest in ..."
In this segment, Goodrow delves into video compression techniques and their relation to understanding video content. He explains how predicting the next frame is akin to compression algorithms and discusses the challenges of maintaining quality while summarizing video information.
"this is the way to go so see you from the perspective of just working with this video how do you think an algorithm that just watches all of YouTube stays up all day and night watching YouTube will be..."
Goodrow reflects on the progress made in video summarization over the past eight years, admitting that the challenge remains largely unsolved. He shares insights on the potential future of YouTube and how it could evolve to better serve viewers and creators.
"compression so the idea is tradition when you think of video image compression you're trying to maintain the same visual quality while reducing the size but if you think of deep learning from a bigger..."
The conversation shifts to the future of YouTube as a platform, with Goodrow envisioning it as a more personalized and enriching alternative to traditional television. He discusses the importance of responsible content curation and the potential for YouTube to enhance viewers' lives through better discovery of content.
"quarter of the way so on that topic what does YouTube look like ten twenty thirty years from now I mean I think that YouTube is evolving to take the place of TV you know I grew up as a kid in the 70s ..."
In the closing segment, Goodrow expresses excitement about the future of YouTube and its role in society. He emphasizes the platform's potential to provide enriching experiences and the importance of ensuring that viewers have access to the best content available.
"you see creators creating visual experiences and virtual worlds so if I'm talking crazy now but sort of virtual reality and entering that space there's that at least for now totally outside of what Yo..."