The promised artificial intelligence revolution requires data. Lots and lots of data. OpenAI and Google have begun using YouTube videos to train their text-based AI models. But what does the YouTube archive actually include?
Our team of digital media researchers at the University of Massachusetts Amherst collected and analyzed random samples of YouTube videos to learn more about that archive. We published an 85-page paper about that dataset and set up a website called TubeStats for researchers and journalists who need basic information about YouTube.
Now, we’re taking a closer look at some of our more surprising findings to better understand how these obscure videos might become part of powerful AI systems. We’ve found that many YouTube videos are meant for personal use or for small groups of people, and a significant proportion were created by children who appear to be under 13.
Bulk of the YouTube iceberg
Most people’s experience of YouTube is algorithmically curated: Up to 70% of the videos users watch are recommended by the site’s algorithms. Recommended videos are typically popular content such as influencer stunts, news clips, explainer videos, travel vlogs and video game reviews, while content that is not recommended languishes in obscurity.
Some YouTube content emulates popular creators or fits into established genres, but much of it is personal: family celebrations, selfies set to music, homework assignments, video game clips without context and kids dancing. The obscure side of YouTube – the vast majority of the estimated 14.8 billion videos created and uploaded to the platform – is poorly understood.
Illuminating this aspect of YouTube – and social media generally – is difficult because big tech companies have become increasingly hostile to researchers.
We’ve found that many videos on YouTube were never meant to be shared widely. We documented thousands of short, personal videos that have few views but high engagement – likes and comments – implying a small but highly engaged audience. These were clearly meant for a small audience of friends and family. Such social uses of YouTube contrast with videos that try to maximize their audience, suggesting another way to use YouTube: as a video-centered social network for small groups.
Other videos seem intended for a different kind of small, fixed audience: recorded classes from pandemic-era virtual instruction, school board meetings and work meetings. While not what most people think of as social uses, they likewise imply that their creators have a different expectation about the audience for the videos than creators of the kind of content people see in their recommendations.
Fuel for the AI machine
It was with this broader understanding that we read The New York Times exposé on how OpenAI and Google turned to YouTube in a race to find new troves of data to train their large language models. An archive of YouTube transcripts makes an…
