Creators: Benedict Hubener, Stephanie Lee, Kelvyn Marte, Andreas Refsgaard and Gene Kogan
The Classifier is an AI paired with a table interface that selects music based on the beverages people are consuming around it at that time. The system is comprised of a microphone that takes in the ambient noise of drinking the beverage and an AI that compares the characteristic sounds to pre-trained models. The current limitation of the models is that they are constrained to three categories: hot beverages, wine or beer. The approach they take in trying to characterize the beverages by their distinctive noises is really unique, yet I have to wonder what other sensors or methods they could’ve used to collect more… telling… kinds of information, such as heat, color, carbonation- which could possibly lead to expanding the beverage categories from the current three.
I’ve also always been fascinated by generating contextual music. In my time at CMU, I’ve come across several student projects who’ve sought to use visual and other sensory input as the lens through which music is created. A friend of mine tried to recreate the sound track of a classic episode of “Tom & Jerry” just purely through a series of frames. Seems like a fascinating and incredibly enigmatic field of study that I would love to pursue.