Abstract: TBA

Bio: TBA

Abstract: TBA

Bio: TBA

Abstract: Modern AI research often pays a high tax for early commitments: once we train, fine-tune, quantize, or deploy a model, changing course becomes expensive. This talk argues for treating model spaces as objects to be navigated, rather than searched blindly. I will discuss three instruments for reducing this tax: inspection, to read trained networks as dynamical systems; inversion, to ask what hidden states still remember; and steering, to reuse expensive robustness improvements as transferable directions. The common theme is extracting more signal from each trained model, turning checkpoints into maps, witnesses, and directions for future decisions.

Bio: Emanuele Rodolà is a Professor of Computer Science at Sapienza University of Rome, where he leads the GLADIA group on AI. His work in this field has been supported by an ERC grant, a FIS grant, and a Google Research Award among others. In the past, he was a postdoctoral researcher at USI Lugano (2016–2017), a Humboldt Fellow at TU Munich (2013–2016), and a JSPS Fellow at the University of Tokyo (2013), in addition to visiting periods at Tel Aviv University, Technion, École Polytechnique, and Stanford. He is a fellow of ELLIS and a fellow of the Young Academy of Europe. Professor Rodolà has received numerous awards for his research and plays an active role in the academic community, serving on program committees and as Area Chair for major conferences in AI and ML. His current research focuses primarily on neural model fusion, representation learning, language models, ML for audio, and multimodal learning, with around 200 publications in these areas. His work has been featured in media outlets including Fortune, Wired, RAI, and Internazionale.

Abstract: Can we learn more from data than existed in the generating process itself? Can new and useful information be constructed from merely applying deterministic transformations to existing data? Can the learnable content in data be evaluated without considering a downstream task? On these questions, Shannon information and Kolmogorov complexity come up nearly empty-handed, in part because they assume observers with unlimited computational capacity and fail to target the useful information content. In this talk we identify and exemplify three seeming paradoxes in information theory: (1) information cannot be increased by deterministic transformations; (2) information is independent of the order of data; (3) likelihood modeling is merely distribution matching. To shed light on the tension between these results and modern practice, and to quantify the value of data, we introduce epiplexity, a formalization of information capturing what computationally bounded observers can learn from data. Epiplexity captures the structural content in data while excluding time-bounded entropy, the random unpredictable content exemplified by pseudorandom number generators and chaotic dynamical systems. With these concepts, we demonstrate how information can be created with computation, how it depends on the ordering of the data, and how likelihood modeling can produce more complex programs than present in the data generating process itself. We also present practical procedures to estimate epiplexity which we show capture differences across data sources, track with downstream performance, and highlight dataset interventions that improve out-of-distribution generalization. In contrast to principles of model selection, epiplexity provides a theoretical foundation for data selection, guiding how to select, generate, or transform data for learning systems.

Bio: Andrew Gordon Wilson is a Professor at the Courant Institute of Mathematical Sciences and Center for Data Science at New York University, and an Amazon Scholar. He aims to develop a prescriptive foundation for intelligent systems. His work includes generalization theory, Bayesian inference, equivariances, time-series forecasting, and scientific applications, particularly in computational biology, physics, and materials. He has received the NSF Career Award, the Heilbronn Distinguished Fellowship, the Amazon Research Award, and several best paper, dissertation, reviewer, and area chair awards.