This must be the best visible example of artificial intelligence on how Youtube recommends your video based on the content of the video (and not only the category or text of the video).
Before we begin: This post will make more sense to an Indian audience because of the content of the example videos I have shown below.
1. Video shown as being played
2. The video recommended
How similar are videos?. If you have gone through both the video fully, you will know that in the video that is shown as being currently played, the actor makes a behavioral mockery of a particular gender. In the video that is being recommended, the same kind of behavioral mockery exists. Both video definitely falls under the light humor genre. Initially, one may not realize how similar both videos are but they are with its content inside.
But it could be cookies doing the trick?
Yes for sure. But not all is done with cookies. Cookies would keep track of videos I have already watched (recommend me episode 2 of web a series) and could at best-recommended videos based on text-similarity. But only an advanced AI engine (deep neural net?) can judge the content of the video in depth and give the recommendation based on the same. You really need to watch both the videos to understand the kind of depth of AI I’m talking about here.
A little digging on the Internet reveals that AI in Youtube has existed for a long time. Probably the training models have now become better over time. A quote from The Verge reads:
Google Brain, the parent company’s artificial intelligence division, which YouTube began using in 2015. . .Brain, however, employs a technique known as unsupervised learning: its algorithms can find relationships between different inputs that software engineers never would have guessed.
“One of the key things it does is it’s able to generalize,” McFadden said. “Whereas before, if I watch this video from a comedian, our recommendations were pretty good at saying, here’s another one just like it. But the Google Brain model figures out other comedians who are similar but not exactly the same — even more adjacent relationships. It’s able to see patterns that are less obvious.”
To name one example: a Brain algorithm began recommending shorter videos for users of the mobile app, and longer videos on YouTube’s TV app. It guessed, correctly, that varying video length by platform would result in higher watch times.
I’m still a novice in AI/ML but I know that Google’s AI neither overfits, nor underfits for sure. 🙂