![\"\"](\"https:\/\/d3.harvard.edu\/platform-digit\/wp-content\/uploads\/sites\/2\/2022\/04\/statistic_id1201880_leading-websites-worldwide-2021-by-monthly-visits.png\")
<\/a><\/figure>\n\n\n\nAs mentioned above, YouTube offers a huge range of content on its platform. Below, you can see the most popular YouTube content by monthly views. T-Series, an Indian music and film company, has the most monthly views with around 3.52 billion views. T-Series immense popularity speaks to the popularity of YouTube in India. As of 2022, India is the country that houses the largest YouTube audience with over 460 million users engaging with the platform (Ceci, 2022). For reference, the United States is second with over 240 million YouTube viewers.<\/p>\n\n\n\n
Closely followed by T-Series is POPS Kids, a leading educational and entertainment provider for kids. The fact that POPS Kids and T-Series are so close in their monthly views (both over 3 billion!) speaks to the breadth and depth of YouTube\u2019s offerings. YouTube is truly a platform with a diversity of offerings. However, the content itself isn\u2019t just the value that YouTube offers, the value also lies in how<\/em> YouTube surfaces and recommends content. <\/p>\n\n\n\n One of YouTubes biggest value offerings beyond the content istslef is the recommendations and curation for suggested videos. When a user first opens up YouTube, they are greeted with a familiar page with new videos. These videos are called recommended based on AI-algorithms. After a user watches a video, they are created with another page of recommended videos. Many users who frequently go on YouTube, including myself, rarely search for videos. Instead, YouTube\u2019s algorithms are so thorough that they are able to drive the viewing experience.<\/p>\n\n\n\n YouTube is able to achieve its terrific recommendations through a two-facing AI algorithm. One component of the algorithm is the candidate generation, and the other component is the ranking network. Candidate generation involves understanding the user\u2019s history and comparing a user\u2019s history with other users\u2019 key information such as how many and the types of videos watched, as well as demographics. The ranking network, \u201cuses a \u2018rich set of features describing the video and user\u2019. This two-tiered system allows the system to handle millions of videos, but also scale down to individual users and provide them with meaningful content\u201d (Anderson, 2019).<\/p>\n\n\n\n The ranking system is particularly important as YouTube receives thousands of new videos uploads every day, and YouTube only wants to recommend high quality videos. To know which videos are high-quality, YouTube measures the time that users spend watching videos as a proxy for the quality of the videos. AI algorithms then use that data to make predictions based on a user\u2019s history how likely they are to watch a video. As Anderson notes, this is achieved through, \u201cweighted logistic regression. The reason it is weighted is that only positive videos are given weight. A positive video is where a user has actually clicked on the video at all, this helps the algorithm learn only about videos that the user interacted with\u201d (Anderson, 2019). Both the candidate generation and ranking network are central to YouTube\u2019s value creation.<\/p>\n\n\n\n YouTube also utilized AI when they rolled out a feature for adding \u2018chapters\u2019 to videos. This novel feature allows users to see clips of videos before watching them. For example, a user would hover their mouse of a video they want to watch and the \u2018chapters\u2019 would begin playing. In a blog post, Jordan from YouTube wrote,<\/p>\n\n\n\n \u201cWe want to make it easier for people to navigate videos with video chapters, so we are experimenting with automatically adding video chapters (so creators don’t have to manually add timestamps). We\u2019ll use machine learning to recognize text in order to auto generate video chapters\u201d (Jordan, 2019).<\/p>\n\n\n\n This machine-learning enabled innovation makes it easier for creators to upload videos without choosing a specific timestamp, and it allows viewers to gain a \u2018sneak peek\u2019 into the content<\/p>\n\n\n\n Nearly 14 years after YouTube\u2019s acquisition, a global pandemic changed YouTube\u2019s operating model and principles. The demand for work-from-home coupled with a rise in misinformation and violence from the pandemic catalyzed YouTube\u2019s innovative use of AI to filter and censor videos. In August of 2020, a few months into the pandemic, YouTube leveraged AI to remove 11 million videos from its site (Narayan, 2020). These videos contained misinformation or violence, and 10.85 million of them were flagged by YouTube\u2019s AI bots. YouTube continues to leverage AI to flag videos for removal.<\/p>\n\n\n\n Overall, YouTube is an increasingly popular platform for video sharing, and AI is central to their operating model. Whether it\u2019s predicting cat videos, teaching kids how to do math, or a \u2018how to video\u2019 about installing gutters, YouTube is the go to place for high quality video content, all thanks to Artificial Intelligence!<\/p>\n\n\n\n References<\/strong> <\/p>\n\n\n\n Ceci, L. (2022, January). YouTube users by country 2022<\/em>. Statista. https:\/\/www.statista.com\/statistics\/280685\/number-of-monthly-unique-youtube-users\/<\/a> DataReportal. (2022, January). Global top websites by monthly visits 2020<\/em>. Statista. https:\/\/www-statista-com.ezp-prod1.hul.harvard.edu\/statistics\/1201880\/most-visited-websites-worldwide\/<\/a> Hosch, W. L. (2022, March). YouTube | History, Founders, & Facts | Britannica<\/em>. https:\/\/www.britannica.com\/topic\/YouTube<\/a> published, R. N. (2020, August 27). YouTube uses artificial intelligence to remove 11 million videos in a quarter<\/em>. TechRadar India. https:\/\/www.techradar.com\/in\/news\/youtube-uses-artificial-intelligence-to-remove-11-million-videos-in-a-quarter<\/a> Rangaiah, M. (n.d.). How is YouTube using Artificial Intelligence? | Analytics Steps<\/em>. Retrieved April 17, 2022, from https:\/\/www.analyticssteps.com\/blogs\/how-youtube-using-artificial-intelligence<\/a> Social Blade. (2022, March). Most viewed YouTube channels 2021<\/em>. Statista. https:\/\/www-statista-com.ezp-prod1.hul.harvard.edu\/statistics\/373729\/most-viewed-youtube-channels\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":" YouTube is increasingly the most popular video sharing platform in the world, all thanks to Artificial Intelligence!<\/p>\n","protected":false},"author":19161,"featured_media":15586,"comment_status":"open","ping_status":"closed","template":"","categories":[],"class_list":["post-15585","hck-submission","type-hck-submission","status-publish","has-post-thumbnail","hentry"],"connected_submission_link":"https:\/\/d3.harvard.edu\/platform-digit\/assignment\/machine-learning-2\/","yoast_head":"\n<\/a><\/figure>\n\n\n\n