The Future of Deep Learning

Ilke Demir depicts the state of generative AI, deepfakes for good, the emotional shelf life of synthesized media, and methods to identify AI-generated content.

Ilke provides a primer on traditional generative models and generative AI. Outlining the fast-evolving capabilities of generative AI, she also notes their current lack of controls and transparency. Ilke then clarifies the term deepfake and highlights applications of ‘deepfakes for good.’

Ilke and Kimberly discuss whether the explosion of generated imagery creates an un-reality that sets ‘perfectly imperfect’ humans up for failure. An effervescent optimist, Ilke makes a compelling case that the true value of photos and art comes from our experiences and memories. She then provides a fascinating tour of emerging techniques to detect and indelibly identify generated media. Last but not least, Ilke affirms the need for greater public literacy and accountability by design.

Ilke Demir is a Sr. Research Scientist at Intel. Her research team focuses on generative models for digitizing the real world, deepfake detection and generation techniques.

A transcript of this episode is here.