D&AD Festival 2019 in collaboration with Shutterstock
This machine is learning
In our hyper-connected world, we use machines to facilitate the sharing of billions of new images every day, but is the machine able to learn independently what is most likely to catch and keep our limited attention? Maŝino Bay’s eye-catching single screen installation uses still images and video clips from the Shutterstock library to facilitate a machine learning project which asks the question ‘What type of imagery is most appealing to the human viewer?’
For our first public facing project we teamed up with Malte Lichtenberg, a PhD student in Machine Learning at Bath University. The machine offered up a selection of imagery via the screen, learning ‘live’ through the reinforcement input signals it received by studying it’s audience. The number of people coming to observe the imagery, and the time their attention was kept by it was recorded by sensors. The data was then fed back to the machine to allow it to measure the preference for specific imagery before proposing new combinations for the next iteration of the algorithm.
Produced in partnership with Greatcoat Films
Special thanks to Jayson Haebich, Luke Todd and Kate Harmatz