Abstract: Extreme events in society and nature, such as pandemic spikes, rogue waves, or structural failures, can have catastrophic consequences. Similarly, uncovering rare optimal designs, such as optimal genetics in crop design or turbulent flow control, can have transformational impacts. Unfortunately, first-principle approaches often cannot characterize these applications in “real-life” as they occur rarely, present unaccountable complexities, are inherently noisy, and can belong to unknown infinite-dimensional systems. Thus, only data, carefully acquired from experiments or high-fidelity simulations, can inherently embed these challenges and provide ground truth insight. This means that decision making for acquiring data, which is finite and expensive in cost and time, becomes paramount for understanding and utilizing the complex systems around us. Here, we present approaches that leverage Machine Learning (ML or AI) for optimal decision making to acquire data that efficiently characterizes complex systems. We show our approach in three contexts, 1) discovering and quantifying extreme events in rogue wave, pandemic spike, and structural failure examples, 2) uncovering optimal designs for turbulent flow control and genetic design of crops, and 3) eliminating misleading data from datasets previously acquired without optimal decision making.  Finally, we conclude by discussing the generality of our AI-Assisted decision-making framework and how these techniques will revolutionize how we interact with, understand, and design complex systems. 

Bio: Dr. Ethan Pickering leads the AI Genomics Platform at Bayer Crop Sciences and is a lecturer in MIT MechE. His research focuses on developing active learning AI models to search complex topologies. At Bayer, these ideas are primarily used to inform genetic design of climate resilient crops. He completed his Ph.D. in Mechanical Engineering at Caltech, developing data-driven reduced-order turbulence models, and a postdoctoral fellowship at MIT with Prof. Themis Sapsis researching extreme and rare event AI.