In the master, I became interested in applying AI to my designs, as this is an increasingly popular technology not always used to its full potential. The courses related to AI taught me the ethics and implementation of Machine Learning, which help me to design for such technology with its limitations in mind.
The project Bird Whisperer (.pdf) (figure 8 & figure 9), of the course Creativity and Aesthetics of Data and AI, explored the interrelations between humans, birds, and AI from a post-human viewpoint. By doing so, it gave me a new perspective on the impact technology and AI can have on daily life and how it can be used for finding new kinds of aesthetics from complex data. The Food Swiping (figure 10) project was part of the Designing with Advanced AI course and applies Machine Learning to offer personalized recipes to help explore new cuisines when uncertain what to eat. Here, I learned how to wirelessly store and access data on a server, which proved valuable for the deployment of Verso (.pdf). Finally, from the course Embodying Intelligent Behaviour in Social Context, StrAIke a pose applies ML to recognize body positions (figure 11) in order to create more dynamic and active learning sessions. All projects implemented a working machine learning model, including a GAN, J48 decision tree, and Naïve Bayes.
The projects made me realize that designs initiating from a technological innovation do not always address actual user needs. I found myself seeking for a problem fitting the ML solution, rather than finding a problem and using ML to solve it. In addition to the courses, I took several TA jobs for the Bachelor courses Making Sense of Sensors, Creative Programming and Data acquisition through embodied sensors. Tutoring required me to solve problems I had never had myself, making me more versatile in data and programming.
I have not developed these skills for the sole purpose or creating a working AI, but instead, I apply my knowledge of the possibilities and limitations of AI as a means to achieve more suitable interactions related to the context-of-use. In other words, for expanding the solution domain and improving my creativity. Since this topic is so broad and quite difficult, I took multiple courses to gain a more complete understanding. In addition, being able to develop a (partly) working AI can be used for testing higher fidelity prototypes.
