Let’s take a quick look at the number one concern people may have with this model, i.e., Privacy. People have started taking digital privacy pretty seriously, and for them, this may come as a shock that their photos are being used to train an AI model.
Combined, Facebook, WhatsApp, Instagram, and Messenger are sitting on a vast treasure trove of data that can be used to provide detailed information about their users across their platforms. Facebook’s data policy clearly states that as long as that user information is posted publicly on their platforms, they can use it to improve their current services and develop newer, better ones, like SEER.
But let’s be honest, none of the users knew that their photos, along with the rest of their data, were being used to serve targeted ads and build AI systems. Looking at some of the current offerings by Facebook, such as its powerful image tagging feature for photos that uses facial recognition shows that the company sees value in understanding their users better and why it is working on SEER.
But like any other bleeding-edge technology, SEER isn’t perfect, and Facebook is worried that the images taken from Instagram may disproportionately represent the younger demographic that has access to the internet and mobile phones and may not be able to accurately represent the groups that are not well-defined by the image dataset.
When asked about the availability of SEER to the rest of the world, Facebook stated that they would offer some part of the software in the form of an open-source library. This would allow other researchers to experiment with self-supervised learning and train their models on uncurated image datasets.