Your Mannequin Challenge Video Is Being Used to Train Robots

An MIT review article says researchers trained AI with 2000 Mannequin Challenge videos that Internet users shared

By any chance, did you participate in the viral mannequin challenge back in 2016? Well, here’s a quick update.

By doing that, you have contributed to advanced robotics research!

Can You Tell Me What Happened?

While as humans, we are able to discern the 3D setup from a 2D video, machines can’t. But as we move into a world of self-driving cars, this knowledge will prove immensely useful to help robots manoeuver in different surroundings.

And so, to tackle that problem - of understanding a 3D scene from a 2D video - to train better networks and robots, a team at Google AI tapped into 2000 videos of people participating in the viral mannequin challenge and efficiently used the repository of free data!

So basically, even though the Mannequin Challenge is no longer trending, it’s videos are aiding advanced science.

Because by making a group of people freeze in one place, and having another person recording a video by moving around and capturing the frozen people from multiple angles with different posing styles, the person recording the video can be thought of as a robot that's trying to make out the gaps and distances between the obstacles.

Shocked that your innocuous video was used like that?

Well, according to MIT Technology Review, which initially reported on the study, it’s pretty common in the AI world, where researchers commonly scrape publicly available images and videos to train bots.

And even though it could be argued that all the videos were available publicly on social media platforms to begin with, the fact that the researchers used videos without using the consent of the people could be unsettling to many, and even raise valid privacy concerns.