Speaker Bio

Dr. Ilke Demir is the Founder and CEO of Cauth AI, building responsible AI platforms for the future of content. Previously at Intel Labs, Dr. Demir single-handedly built the Trusted Media research stream, to combat misinformation and protect artists’ rights using human-centered AI research, releasing the world’s first real-time deepfake detection platform. Her honors include Jack Dangermond Award, Bilsland Dissertation Fellowship, DEG EnTech Awards finalist, and Cannes Lions Gold. Recognized as an ACM Distinguished Speaker and IEEE Industry Distinguished Lecturer, she has delivered over 150 keynotes, panels, and talks worldwide, and served multiple times as a US gov. generative AI trainer. Her doctoral work on proceduralization of 3D data, at the intersection of generative representations and geometric/topological shape processing, pioneered generative model research over a decade before it became mainstream. As a postdoc at Facebook, she developed revolutionary generative street addresses helping four billion people lacking proper addressing systems. Her technical contributions at Pixar Animation Studios enhanced beloved films including Finding Dory, Coco, and Toy Story 4, while at Intel Studios she established the research foundations for the world’s largest volumetric capture stage, bridging creative processes with AI innovation. In addition to her publications in top-tier venues (SIGGRAPH, ICCV, CVPR), she has organized workshops, competitions, and courses in deep learning, computer vision, and graphics (DeepGlobe, EarthVision, OpenEDS, USM3D, etc.). Her work receives extensive coverage from CNN, BBC, MIT Tech Review, The Economist, and PBS Nova. With multiple patents protecting critical innovations, Dr. Demir shapes AI policy conversations at the World Economic Forum while contributing to global standards through C2PA and other organizations, positioning her as both creator and guardian of humanity’s digital future.

Title: Can You Still Trust the Eye in the Sky?
Abstract: Satellite imagery underpins critical decisions in national security, environmental monitoring, and urban planning; yet advances in generative AI now make it possible to fabricate convincing satellite images with little more than a consumer GPU. This keynote traces the full lifecycle of the problem, from early computer vision frameworks for analyzing and synthesizing geospatial data to fake satellite imagery nowadays being weaponized in war misinformation. Next, we explore synthesis for good timeline where generative models are creating real-world impact. On the defense side, we survey highly effective detection methods, alongside proactive protection strategies including provenance tracking and watermarking. We argue that safeguarding geographic truth demands a layered approach, combining passive forensic analysis with active protection embedding. We end the talk with open challenges not yet applied or solved in geo-space like adversarial reality, human perception, geospatial data literacy, and multispectral complexity.