New Tools Teach How To Detect Deep Fakes

·53 min·3 clips

My Art, My Choice turns an image into a protected version that breaks diffusion models like DALL-E and Stable Diffusion.

Don Murphy introduces the episode as a look at the battle over deepfakes. He sets the stakes right away, saying it is getting harder to tell what is real from what is fake, especially on social media and even on TV networks. Dr. Ilka Demir joins to explain the field from the inside. Her background spans 3D vision, computational geometry, telemetry, generative models, remote sensing, and deep learning. She describes how hard it has become to tell real media from synthetic media. The discussion treats that problem as a public one, not just a technical one. Demir explains that detection systems can be used across many domains. She points to collaborations with nonprofits and human rights organizations. Those groups try to identify high impact, high risk deepfakes before they spread widely. The conversation notes that political deepfakes have already appeared in countries including India and Georgia. Those cases show how synthetic media can be used to sway public opinion. Demir says journalists and news organizations often rely on outside detection teams to help catch material early. That timing matters because a fake can stay small or go viral depending on when it is flagged. The episode also looks at social media platforms as a major place to deploy detection tools. Demir argues that detection should be part of every video, media, image, and audio consumption platform. She then shifts to provenance. Provenance means knowing who created a piece of media and how it was made. It also means knowing whether it was created with consent. It can include which generative model was used. The episode ends with a broad case for transparency around synthetic media. The core idea is simple. If people are going to live with deepfakes, they need better detection and better disclosure.

As heard by us

A grounded look at deepfake creation, detection, and provenance.

Deepfakes are the subject, but the larger concern is how to judge what can still be trusted when images, audio, and video are easy to forge. The episode works best when it treats detection as part of platform design, newsroom routines, and partnerships with nonprofits, human…

Read the full review in PlayNext →

Why you'd press play

You want a clear guide to spotting fake media before it fools your feed.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Nature and Science Podcast