Photo Credit: Pixabay
As part of its new “Learning From Videos” project, Facebook is training artificial intelligence (AI) on videos already uploaded to its platform in order to understand you better. What could possibly go wrong? The same company that has been profiting from user data for more than a decade is now taking it one step further. Facebook has already trained AI on billions of images from Instagram (which it owns), and is now training AI on millions of public videos on its platform. Of course, nobody has really consented to this, and Facebook has been downplaying what exactly it plans to do with its new AI tools, probably to avoid any type of public outcry. But, nonetheless, alarm bells should be ringing right now.
In the base-case scenario, Facebook’s aims and goals for the new AI program sound fairly benign, if not downright boring. The goal, says Facebook, is simply to develop new captioning tools for videos, new content moderation tools, and new functionality to tell who or what is in each video. And, given Facebook’s huge global presence, AI might one day be able to help Facebook distinguish between different visual cues and nuances across cultures and regions. (A hand gesture in one culture or region might mean something very different someplace else around the globe.) OK, fair enough, this is not exactly any reason to get concerned about the doings of Mr. Zuckerberg.
But let’s turn things up a bit and move to Scenario 2. In this example, Facebook would use its advanced AI system to develop better and more robust search functionality. One example that Facebook specifically mentioned is a “digital memories” concept that actually sounds somewhat endearing – a Facebook user missing his or her grandma might one day be able to do a search along the lines of, “Show me all the videos where we sang Happy Birthday to grandma.” Then, thanks to the magic of AI, Facebook would be able to present a complete playlist of grandma birthday videos. Or, Facebook might also be able to develop better content recommendation systems. So, for example, once Facebook learns that you enjoyed that one cute kitten video from your friend, maybe you’ll like this other cute kitten video…
But do you really think that Facebook is going to stop there? Remember, Facebook’s entire business model is built around advertising. So it’s only natural to assume that Facebook will somehow use new AI tools to learn more about your beliefs, preferences, hobbies and interests. By watching a few videos from you, for example, Facebook might get some great insights into what makes you tick as a consumer. In the old pre-digital days, it would be like a bunch of marketing executives watching from behind a two-way mirror as a focus group discusses what they like and what they don’t like. But that takes time and money, right? If you have millions of videos, that’s a much richer source of data. And Facebook is really speeding up how fast AI “learns,” so that it takes a very short period of time to analyze all those videos.
Once the AI learning has reached a certain stage, Facebook would be able use all that data for super-specific ad targeting. If you post a workout video, for example, Facebook might be able to alert its advertisers that you might be in the mood to be pitched some new workout clothes or a pair of new sneakers. If you post a foodie video, you might get restaurant ads and food company ads in your newsfeed. The potential here is almost unlimited. AI would be able to parse nearly any video to come up with advertising suggestions. Imagine how much advertisers might be willing to pay for that sort of data and information.
So the next time you upload a public video to Facebook, just remember that friendly little AI bots will be watching – not just your friends and family. Of course, you’re not getting paid a dime by Facebook for this. The only one getting paid in all this will be Facebook, and that’s why we all should be more than a little bit concerned that Facebook is training AI on public videos of its users. It represents just the latest step in Facebook’s business model of profiting from the data of others.