As computer science and machine learning has progressed, machines are able to make better predictions. However, they are not perfect and most likely won’t be in the near future. To develop on a point made in class, the mistakes that machines make may be the most interesting to examine.
Natural language processing is a developing field that takes from machine learning methods. One potential application of NLP is processing historical financial data (news in text) to predict how a company will do financially in the future. NLP of the news can be combined with regression on numerical financial data to increase accuracy and precision.
- INPUT: financial data (present and historical) both news (text) and stock prices (numbers)
- BLACK BOX: NLP and ML regression to predict financial numbers of future stock numbers, generative color and sounds
- OUTPUT: projector (morphing stream of colors) and speakers
This installation compares the predictions to the actual results in real-time and affects a stream of colors and sound. When the predictions are correct, the colors and sounds display harmonious patterns, the algorithms generate harmonies that are concordant and the colors flow smoothly. However, when the predictions are incorrect, discordant sounds are played that are seemingly random and the colors disperse and move in unnatural patterns. This is a continuous process that results in a stream of sounds and colors that respond to the predictability (and unpredictability) of the financial world. The implementation of the sound is possible using MIDI and generative music algorithms while entropy can be conveyed through random sound triggers.