Raw LLM Responses

My take as an engineer who works in data science/ML - facebook **already owns** 10+ years of photos from their users, and they can already tag faces. It would be much easier for them to query the photos they already own and use that to train, than to spend all the time and money to start a viral campaign to trick users into doing it for them. Why would they do it the hard way? Steps to build data set through 10 year challenge: 1. Start a viral campaign 2. Wait for users to upload sufficient data 3. Automate the process of splitting the composite images into 2 images of faces OR automate cropping the faces out of the composite image Steps to build with querying: 1. Query for all images that are 10 years apart & & have a recognizable face & & are tagged with a specific user Building a data set with the 10 year challenge is slower, more expensive, less reliable and harder to control. It just doesn't make sense for them to collect data this way; **they already have it**!