Case Study: AI Tinder Bot
We trained an AI that learns preferences and swiped Tinder
We asked ourselves: Can an AI learn which people you find attractive and thus automate profile matching?
Profile swiping on Tinder is a rather repetitive task: Based on a few pictures and some text, you decide whether you like the person enough to give a positive swipe and potentially match. A task for which an AI would be suitable! So we set about training a Deep Convolutional Neural Network, which learns which profiles you particularly like based on a few hundred Tinder swipes. The AI then analyses the image and outputs an attractiveness score - personalised to the person the AI has trained. With a bit of cunning and trickery, we managed to link the AI to the Tinder database without having to use the Tinder iOS/Android app. After a few days, we had not only successfully automated thousands of swipes, but also received enquiries from international media reporting on our project. We open-sourced the entire source code and showed the world the possibilities but also the dangers of AI.
Learns subjective attractiveness from images
Published source code
International press attention