Schedule: Interacting with AI sessions

At the core of today's problems with image classification and deep learning lies one fundamental truth: most AI systems operate by choosing the path of least resistance, not the path of highest long-term quality. Matt Zeiler discusses Clarifai's approach to closing the loop on AI and the techniques it employs to counter the AI quality regression phenomenon. Read more.

Digital character interaction is hard to fake, whether it’s between two characters, between users and characters, or between a character and its environment. Nevertheless, interaction is central to building immersive XR experiences, robotic simulation, and user-driven entertainment. Kevin He explains how to use physical simulation and machine learning to create interactive character technology. Read more.

Games are wonderful contained problem spaces, making them great places to explore AI—even if you're not a game developer. Paris Buttfield-Addison, Mars Geldard, and Tim Nugent teach you how to use Unity to train, explore, and manipulate intelligent agents that learn. You'll train a quadruped to walk, then train it to explore, fetch, and manipulate the world. Read more.

Following the launch of CTRL-labs’s developer kit, CTRL-kit (neural interface device), Patrick Kaifosh paints a picture of a world with neural interfaces, explaining how this technology will change our lives. Patrick outlines a future where we'll be looking up at the world instead of down at our phones—and leads a live demo of CTRL-kit in action. Read more.

What's it like to be a mobile phone or to attach a wind sensor to a neural network? Jeff Thompson outlines several recent creative projects that push the tools of AI in new directions. Part technical discussion and part case study for embedding artists in technical institutions, this talk explores the ways that artists and scientists can collaborate to expand the ways that AI can be used. Read more.

Forough Poursabzi-Sangdeh argues that to understand interpretability, we need to bring humans in the loop and run human-subject experiments. She describes a set of controlled user experiments in which researchers manipulated various design factors in models that are commonly thought to make them more or less interpretable and measured their influence on users’ behavior. Read more.

Artificial intelligence has shown great potential to revolutionize clinical medicine and healthcare delivery. However, incorporating these algorithms into clinical workflows involves a big challenge: convincing clinicians and regulators to trust a “black box” solution. Behrooz Hashemian explains how he's helping make deep neural networks interpretable to provide evidence for clinical decisions. Read more.

Humayun Irshad offers an overview of an active learning framework that uses a crowdsourcing approach to solve parking sign recognition—a real-world problem in transportation and autonomous driving for which a large amount of unlabeled data is available. The solution generates an accurate model, quickly and cost-effectively, despite the unevenness of the data. Read more.

Andrew Caosun discusses a framework that unifies hidden Markov models and deep learn algorithms (RNN) with modeling components that consider long-term memory and semantics of music (LSTM and convolution). It takes users' original creations as input, modifies the raw scores, and generates musically appropriate melodies. Read more.