Design Intelligence is a new subject that introduces students to a practical, hands-on approach to machine learning and artificial intelligence. Providing a new lens through which to engage machine learning through aesthetic, form-finding and interaction, the course introduces students to neural networks, CNNs, RNNs, GANs and large-scale generative language models, as well as how to collect and prepare data for training their own neural networks. Situated within a graphic, product and interaction design context, students learn to develop a new kind of creative practice that not only actively engages in shaping the future of artificial intelligence, but is also instrumental in addressing its biases and failures.
The course is divided into two parts. In the first half, students progress through a series of 4 short warm-up exercises that give exposure and hands-on experience to different neural network architectures and techniques. In the second half, students develop an independent project, further exploring ideas uncovered during the exercises or pursuing their own interests.
Exercise 1: From Parametric to GANs
Students create parametric drawings in p5js that act as input for generative adversarial networks, experimenting with and comparing the trade offs of both generative methods.
Exercise 2: Interactive Drawing Machine
Students develop an interactive drawing machine by creating a unique dataset, training a classifier, and using its output as a source of dynamic input for visual composition and design.
Exercise 3: Neural Fabricator
Students generate synthetic data in order to train a variational autoencoder. Outputs from their neural network acts as source material for 3D design and fabrication.
Exercise 4: Music Video
In the last exercise, students create a music video by combining VQGAN + CLIP and a recursive neural network (RNN) for audio and sound generation, bringing together a range of different neural networks architectures in order to compose a single visual experience.
Inspired by the concepts and techniques seen earlier in the course, students develop a longer and more in depth project, pursuing their own personal interests in art, design, interaction, artificial intelligence, and neural networks.
COPYRIGHT © 2024