Gennady Kogan

This course is a survey of some of the more exploratory new directions AI is heading towards in the coming years and the creative applications these new developments may enable. More specifically, the class will explore the following three subtopics:

1) Realistic language models: Turing-test passing text generators like GPT-3 are writing whole paragraphs with human-level coherence. We will explore techniques for generating creative fiction and chatbots using both APIs and our own home-brewed NLP models.

2) Generative music: Since WaveNet in 2016, generative models of audio have gradually evolved to capture music and natural sound. How and for whom will top-10 hits in the future be made, and how will they be listened to? We will speculate about such questions and experiment with speech and audio synthesizers and new kinds of digital musical instruments.

3) Decentralized AI: What happens when we can do machine learning at scale without putting all the data in one computer? This seemingly trivial difference opens up all kinds of new possibilities, including AI co-ops and autonomous artificial artists.

Each subtopic is accompanied by practical resources to use or implement these techniques for yourself, as well as an overview of the available and relevant tools associated with it, and will finish with a short project to apply them to a topic you’re interested in.