Most people are well aware of the fact that people and computers are becoming socially bound. Talk to any person and it’s almost certain that their smartphone or computer will factor in somehow. What’s less well known is how much easier the human side of things has been. Humans relate to technology far easier than technology relates to humans. One of the biggest reasons why comes down to how the human brain operates. People unconsciously use quite a bit of brainpower to recognize human faces. The sheer difficulty of recognizing a face can be seen by looking at non-human primates. They’re the closest relatives of humanity, and have fairly humanlike features. But most people would have to really work to distinguish the faces of multiple shaved chimpanzees from each other. All that hard work distinguishing small differences is normally handled by the human unconscious as people look at different faces. But it’s been as difficult for machines to tell humans apart as it would be for a human to tell chimpanzees apart.
All of this might be changing thanks to a recent breakthrough with the tinder dating app. Stuart Colianni has combined information from the visual components of tinder with Kaggle. The earlier information is something people need to keep in mind when setting their expectations for the project.
Humans are quite literally incapable of really understanding how much of their neural architecture goes into facial recognition. This makes the earliest hope for this work seem rather modest. The initial projects using the combined tinder data will be used to help create artificial learning systems capable of differentiating between male and female faces. But this seemingly modest task is something that takes a great deal of neural activity for humans. If machines can get a boost from this project than it’ll be a huge step forward for machine learning systems and artificial intelligence.
Tinder has yet to publically comment on the project. But in general they’ve been fairly friendly to 3rd party use of their data. Tinder’s API has made it fairly easy to implement bridges to any 3rd party code. Over the years, Tinder has generally been tolerant of projects which do so. Given that this machine learning project is fairly modest in terms of direct user interaction it seems probable that they’ll have little problem with it either.