Squares is a research project which explores the idea of technology having its own desires and learning its own tastes by using data from the Foursquare API, input from a human, and a neural network.
Squares are social organisms that need people to move them around. They develop preferences based on venue data from the foursquare api, and how much 'fun' that square had at the venue. This data is used to train a neural network (venue data as inputs, 'fun' score as output) after each outing.
When a user logs in, the neural net is executed on all nearby venues, which leads to the user's square suggesting places it wants to go. For instance, mine really likes the Key Foods near my apartment.
'Fun' is determined in a simplistic way, depending on how much the user interacts with the square. In the future, we'd very much like to add a way for squares to interact with one another, to add another layer of depth.
Fred Truman and I collaborated on Squares, and Fred went on to explore some of these concepts in his ITP "thesis" project.
Intel funded a research group at ITP to explore "Vibrant Technology", which, very simply put, is the idea of viewing technology not as a tool to benefit people, but as a peer, with it's own desires. The group was lead by Heather Dewey-Hagborg from ITP and Maria Bezaitis from Intel.
- The backend is written in python using django
- For the neural nets we're using FANN which is great, though I wouldn't recommend the python bindings. It's pretty hacky. Though I'm no a Swig expert.
- Foursquare API
Get a Square and try it at totalsquares.com