An Interview with Ilke Demir, Senior Research Scientist, Intel
I first realized I was interested in artificial intelligence as a possible career path when I was still in my undergraduate studies. My first internship was in a robotics lab, and I was astonished by the interaction between machines and the way computer vision helped robots “see.” At that moment, I knew I wanted to work in computer vision and graphics. Years later, due to my research during my PhD in Computer Science and hands-on experience in deep learning methods during my postdoc and other companies, I gained a holistic understanding of how AI applications work in production. Not long after that, I joined Intel.
My passion in AI includes computer vision and deep learning methods. Of course, when someone says they’re interested in a career in artificial intelligence, it could mean any number of things. First, it probably means they share the same advantage—or defect?—that I have, where you see the real world through the lens of distributions and data (funny how when you’re interested in making machines think more like you, you end up thinking more like a machine). But AI is a large umbrella term that involves many different research domains—from robotics to computer vision to machine learning and beyond.
Before my interview with Intel, I assumed it was just a chip company. So, when I walked in for my interview and saw the types of technology being leveraged for volumetric capture, everything clicked. It was like all my education and research from the time I was an undergrad student through my postdoc and beyond came together for me at Intel. It was a perfect storm of AI innovation paired with a diversity of projects with which I could create my research heaven and impact the whole entertainment industry.
During any given film or video shoot, we capture 270 gigabytes of data per second. That’s a lot of data, but our server farm has petabytes of storage on-site! The only way to process all that information is to go beyond automation and use AI tools to infer real-world representations on high quality 2D and 3D production data. Which means we’re not just writing scripts—we’re making the machines learn by themselves.
Another area of AI that I am particularly interested in is the concept of deep fakes, where AI is used to generate photorealistic videos of someone doing something that never occurred, because it has been fabricated using AI. This has much larger implications than just bad press. If they are mistaken for real videos, it could result in legal, political, financial, and personal consequences. So, one branch of my current research investigates leveraging machine learning on biological signals and other tells to identify deep fakes. We can use AI to track deep fakes by their heart beats!
Before I came to Intel, I had a background in 3D vision, inverse proceduralization, and deep learning. I was also seasoned on converting research prototypes to send them into production. Although prior productization pipelines involved multiple teams, being a part of the end-to-end creation was more abstract. At Intel, I get to be hands-on with real AR/VR productions, working with the VFX team, the actors, and the pre-production team. Really, we exist at the intersection of art and science.
All these AI tools my team is innovating are enabling us to build novel realities—virtual worlds, augmented worlds. We’re reducing the physical limits of filmmaking. We can put objects in post-capture that don’t exist in the real world. We can simulate as many cameras as we want, in as many places as we choose. At Intel, we’re creating immersive experiences within these novel worlds. And in doing so, we’re introducing the new dimension in filmmaking. And no matter where you fall on that spectrum—whether you’re a user, or a creative; an engineer, or an AI scientist like me—the work we’re doing will allow you to be immersed in novel experiences of those digital worlds we create.
Interested in a career in AI?