New stream: TENTATIVE START DATE SPRING 2023

Full Stream Name: Creative Artificial Intelligence

Research Educator: Uli Grasemann

Principal Investigator: Risto Miikkulainen

Credit Options: Spring & Fall

The past few years have seen substantial advances in the ability of computers to create novel and realistic artifacts. Recent AI systems can create photorealistic images from sketches or descriptions, or create convincing pictures of people who do not exist. They can solve difficult design problems, often in unexpected ways that we'd certainly call creative if a human did it. They can produce coherent text on any topic, in the style of an author you choose, and their output is often indistinguishable from language produced by humans. In this stream, we will explore how these and other systems work, and you will design and build your own 'creative' AI application. Starting with basic techniques, we will build up to the algorithms behind state-of-the-art AI systems like GPT-3 and DALL-E. We will mostly focus on Deep Learning (i.e. modern neural networks) and Evolutionary Algorithms (which use the mechanics of natural evolution to solve computational problems).  Along the way, we will utilize a range of useful tools like Python notebooks, scikit-learn and TensorFlow. Each student will choose a specific  AI technique or application to explore in depth, and give a presentation about it in class. In the second part of the stream, students will work in teams to come up with an idea for an interesting project, and then work on turning their idea into a working AI system.