Ali Farhadi

Farhadi earned his doctorate at the University of Illinois Urbana-Champaign under David Forsyth, then spent a year as a postdoctoral fellow at Carnegie Mellon’s Robotics Institute before joining the University of Washington. AI2 announced his CEO appointment effective July 31, 2023. His current faculty biography identifies that leadership role as former and lists his Microsoft Superintelligence position.[1][4]

From visual research to company leadership

Farhadi joined AI2 in 2015 to establish its computer-vision team, researching how actions and interactions help a system understand a scene. He co-founded the on-device deep-learning company Xnor.ai, which Apple acquired in 2020, and subsequently led machine-learning efforts at Apple. His return to AI2 as CEO followed that period.[4]

XNOR-Net, coauthored with Mohammad Rastegari, Vicente Ordonez and Joseph Redmon, studies approximations to convolutional neural networks. One version stores filters as binary values; another also binarizes the input to convolutional layers so much of the computation can use binary operations. The experiments measure the resulting trade-off between ImageNet classification accuracy, memory and computation. This is a concrete route toward less expensive vision models, with accuracy and implementation constraints still relevant.[3]

Farhadi’s research interests include the relationship between language and vision, visual reasoning and how semantic information helps a system understand an image. His university research group provides a way to follow the papers behind those themes, beyond changes in executive titles.[1][2]