July 18th, 2019

FaceApp: Catchy Title

Most likely I do not have to explain what the Faceapp is (or I do (which maybe is a good thing)) but just to get everyone working with the same dataset, here are the basic facts:

In January of 2017 company based in St. Petersburg released FaceApp, an app that let's users manipulate photos using the companies AI software ~~> hairy old selfies. To understand how this process works requires more knowledge of machine learning than I have is necessary, but essentially you open the app and allow access to your camera roll wherein the software within your phone begins to run a program that roughly identifies faces. Then you either take or select a photo, which is uploaded to servers around the world, mostly in the U.S. and Australia - not Russia, owned by Amazon - not Russia. Then software identifies the elements of your face. This is accomplished by showing what is called a neural network images of faces and having it guess where the mouth, nose, eyes, are. The developer then rates (sort of) the success of this guess, and the machine adjusts its algorithm to be more accurate. This is the same process used for modifying the face, if we want it to smile then the machine attempts to create a hybrid of what it has been trained to see as a smile and what it has been trained to identify in any photo as the mouth. The developer rates the hybrid (again, sort of) and the machine becomes better at creating the approximation. 

What is the important takeaway? 

1. When using the app only photos that are selected are uploaded. 

2. None of the photos reach Russian servers 

3. more faces = better algorithm i.e. Faceapp will learn much much faster and better if shown 10,000,000 faces than it will if shown 1,000 faces. 

Now, as you probably know, the app has had a resurgence from (what was most likely a paid marketing push) the “face app challenge” where celebrities use the app to age or beardify their faces. Upon release their had been controversy surrounding a “Hot” feature which because of “biased data sets” (white people) had lightened the skin of the photo. Now, following a few tweets containing quoted elements of the terms of service:

 You grant FaceApp a perpetual, irrevocable, nonexclusive, royalty-free, worldwide, fully-paid, transferable sub-licensable license to use, reproduce, modify, adapt, publish, translate, create derivative works from, distribute, publicly perform and display your User Content and any name, username or likeness provided in connection with your User Content in all media formats and channels now known or later developed, without compensation to you. When you post or otherwise share User Content on or through our Services, you understand that your User Content and any associated information (such as your [username], location or profile photo) will be visible to the public.