COLLOQUIUM: Aleksandra Faust, "The Synthetic Flywheel: Self-Improvement and Simulation in Foundation Models"

Refreshments Provided.

Abstract: 
The journey toward real-world autonomy in AI is fundamentally driven by the mastery of simulated environments and algorithmic self-improvement. This talk explores the evolution of agent training, tracing its roots in classical simulators where "learning to learn"—via automated curricula and AutoRL—first emerged as a critical lever. As the field transitioned to data-driven world models, these early meta-learning principles set the stage for today's frontier: generative foundation models that bootstrap their own capabilities. Next, we will examine how agents actively iterate and advance through self-improvement loops. We will dive into mechanisms like many-shot in-context learning, autonomous self-correction, and long-horizon reinforcement learning that allow models to escape the limits of static training data. By exploring applications across diverse domains—from web agents navigating dynamic digital interfaces, to foundation models achieving sub-angstrom precision in molecular design, to clinical agents mastering diagnostic reasoning via simulated patient encounters (ResidencyRL)—we will see how synthetic data and self-correction act as the engine of modern AI. Finally, by mapping these accelerating capabilities against levels of Artificial General Intelligence, we will analyze the critical interplay between performance, generality, and safe autonomy as we deploy agentic AI into high-consequence realities.

Bio:
Aleksandra Faust is a Director of Research at Google DeepMind, where she leads Frontier AI Health efforts. Her research focuses on foundation models and world models for complex adaptive systems, treating the AI design pipeline as a learnable, sequential, and self-improving decision-making process. This methodology has driven state-of-the-art improvements across drug discovery, robotics, autonomous driving, and web agents, and led to her founding the field of Automated Reinforcement Learning (AutoRL). Notably, she co-authored the seminal "Levels of AGI" framework and led the Gemini Self-improvement research team, developing the reinforcement learning methods behind the Gemini model family. Previously, Aleksandra served as Chief AI Officer at Genesis Molecular AI and held foundational leadership roles at Google Brain, Google Robotics, and Waymo/X. Earlier in her career, she was a Senior R&D Engineer at Sandia National Laboratories. Faust holds a Ph.D. in Computer Science with distinction from the University of New Mexico and an M.S. from the University of Illinois at Urbana-Champaign. She is an IEEE Fellow and a recipient of the IEEE RAS Early Career Award for Industry and the Tom L. Popejoy Dissertation Award, and was named a Distinguished Alumna of the UNM School of Engineering. Her work has been featured in The New York Times, The Economist, and Forbes, and has received multiple Best Paper Awards at premier robotics, machine learning, and systems architecture venues.

Part of the Siebel School Speakers Series. Faculty Host: Nancy Amato

Meeting ID: 845 5733 4528 
Passcode: csillinois

If accommodation is required, please email <erink@illinois.edu> or <communications@cs.illinois.edu>. Someone from our staff will contact you to discuss your specific needs