When babies watch objects fall, they don’t memorize the trajectory of...
The CDS Monthly Research Feature |
When babies watch objects fall, they don’t memorize the trajectory of every dust mote in the air — they learn that things drop when released, developing what scientists call intuitive physics through selective attention to what matters. CDS founding director and Professor Yann LeCun has spent years arguing that this ability to ignore irrelevant details while grasping essential patterns is what separates biological intelligence from today’s data-hungry AI systems, and his team’s new V-JEPA 2 model demonstrates this principle at unprecedented scale.
|
|
|
People are already using chatbots for mental health support without any scientific understanding of the risks involved. Now CDS Associate Professor of Linguistics and Data Science Tal Linzen and CDS Assistant Professor of Psychology and Data Science Grace Lindsay have been awarded funding as part of a new $20 million National Science Foundation institute aimed at combining research on human and machine cognition, with the goal of creating AI systems that can interpret a person’s unique behavioral needs and provide helpful feedback in real time.
|
Machine learning’s most puzzling phenomena — from models that memorize random data yet still generalize well when trained on structured data, to networks that perform better with more parameters than training examples — can be explained by mathematical frameworks developed decades ago. Moreover, these phenomena can be intuitively understood, and reproduced using models as simple as a basic polynomial. CDS and Courant Professor of Computer Science and Data Science Andrew Gordon Wilson argues in a new ICML paper that challenges the widespread belief that deep learning requires rethinking how we understand machine learning.
|
Government agencies rushing to deploy sophisticated prediction systems to identify vulnerable populations may be missing a simpler solution: hiring more staff. New research by incoming (Fall ’26) CDS and Courant Assistant Professor Juan Carlos Perdomo and collaborators reveals that expanding bureaucratic capacity often delivers greater improvements in helping those most in need than incrementally refining algorithmic models.
|
CDS is seeking innovative researchers and passionate educators to join our team. For those pushing the boundaries of data science research, our prestigious Faculty Fellow positions offer the opportunity to lead cutting-edge projects in areas like machine learning, responsible AI, and natural language processing. For educators passionate about shaping the next generation of data scientists, we have an opening for Clinical Faculty focused on teaching and curriculum development.
|
|
|
Copyright (C) 2025 Center for Data Science. All rights reserved.
|
|
|
|
