Going Deep on Deepfakes (feat. Hany Farid)
The FAIK Files
Semantic Guardrails and Training Data Controls
Hany recommends input/output moderation, training-set restrictions, and ongoing red teams to prevent harmful outputs.
Welcome back to The FAIK Files!
In this week's episode:
- We sit down with deepfake expert Hany Farid to discuss the real-world harms of synthetic media.
- Exploring the physics of deepfake detection and why real-time streams might be easier to defend.
- The dangers of using AI to "enhance" images and hallucinate hidden details.
- A look at solutions like C2PA, watermarking, and the pressing need for platform accountability.
Check out Hany's company, Get Real Security here: https://getrealsecurity.com
Learn more about C2PA at: https://contentcredentials.org
Want to leave us a voicemail? Here's the magic link to do just that: https://sayhi.chat/FAIK
You can also join our Discord server here: https://faik.to/discord
*** NOTES AND REFERENCES ***
What Keeps Hany Farid Up at Night?:
- The rising harms of non-consensual intimate imagery (NCII) and child sexual abuse material (CSAM) generated by AI.
- Voice cloning being weaponized for individual fraud and real-time deepfakes used by state-sponsored actors.
- Why the specific tool (like Sora or face swap) matters less than the overall threat vector and resulting harm.
Deepfake Detection - APIs vs. Physics:
- Hany's work at UC Berkeley and his company Get Real Security.
- Why detecting real-time manipulated video is actually easier than identifying well-crafted, file-based deepfakes online.
- How physical camera imperfections (noise) differ from the artifacts introduced by AI upsampling and diffusion models.
The Danger of AI "Enhancement":
- Why using AI to "remove a ski mask" or enhance low-res footage is not like CSI—it's just hallucinating statistically consistent pixels.
- AI lacks a notion of uncertainty, leading to dangerous misidentifications and real-world harm.
- The "Liar's Dividend": When the flood of AI slop makes people doubt the authenticity of real, unedited evidence.
Safeguards, C2PA, and Platform Responsibility:
- The role of watermarks and C2PA content credentials acting as "nutrition labels" for digital media.
- Clarifying that C2PA relies on signed credentials and a trust list, not the blockchain.
- The dire need for social media platforms to enforce semantic guardrails and take responsibility for the content they amplify.
- Find more of Hany's work by searching YouTube for his lectures on "Physics-Based Photo Forensics."
*** THE BOILERPLATE ***
About The FAIK Files:
The FAIK Files is an offshoot project from Perry Carpenter's most recent book, FAIK: A Practical Guide to Living in a World of Deepfakes, Disinformation, and AI-Generated Deceptions.
- Get the Book: FAIK: A Practical Guide to Living in a World of Deepfakes, Disinformation, and AI-Generated Deceptions (Amazon Associates link)
- Check out the website for more info: https://thisbookisfaik.com
Check out Perry & Mason's other show, the Digital Folklore Podcast:
- Apple Podcasts: https://podcasts.apple.com/us/podcast/digital-folklore/id1657374458
- Spotify: https://open.spotify.com/show/2v1BelkrbSRSkHEP4cYffj?si=u4XTTY4pR4qEqh5zMNSVQA
Want to connect with us? Here's how:
Connect with Perry:
- Perry on LinkedIn: https://www.linkedin.com/in/perrycarpenter
- Perry on X: https://x.com/perrycarpenter
- Perry on BlueSky: https://bsky.app/profile/perrycarpenter.bsky.social
Connect with Mason:
- Mason on LinkedIn: https://www.linkedin.com/in/mason-amadeus-a853a7242/
- Mason on BlueSky: https://bsky.app/profile/wickedinterest.ing
Learn more about your ad choices. Visit megaphone.fm/adchoices
The AI-powered Podcast Player
