Defending our reality in the era of deepfakes
summary
Describes concrete defenses and limits: AI Foundation builds the Reality Defender (rd2020.org) platform that aggregates detectors and returns confidence scores; detection uses pixel‑level statistics, background inconsistencies, lip‑sync correlation and voice signatures; Facebook/Kaggle style contests and Microsoft funding accelerate models. Gives sharp examples: GPT‑3 essays fooled judges, an app (ZAO) democratized face‑swapping, and a corpus of 100,000 fake faces is freely available. Warns detection is a moving target and long‑term solution likely requires cryptographic authentication.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 2:41Perceptual Intelligence
- 5:15Big Data AI
- 7:40Uses of Synthetic Reality
- 8:56Deepfake AI
- 11:07Perception is reality
- 11:41AI controls perception
- 12:40Essay writing competition
- 13:02Fake news
- 13:37Guardian oped
- 13:58Fake faces
- 14:22Zhao
- 15:20Gal Gadot
- 15:46Deepfake laws
- 16:34AI part of the solution
- 17:01Detecting imperfections
- 17:34Examples of imperfections
- 19:04Crowdsource deepfake detection
- 20:00Election manipulation
- 23:14RD2020
- 23:24Moving target
- 25:16The most vulnerable time
- 25:58Summary
related talks
26mDeepfakes, audiofakes, and the future of media
30mReality Under Attack: Just How Serious Is the Deep Fake Threat?
45mFUD vs. Reality: Bracing for an AI-Powered Threat Landscape
30mAI-Generated Synthetic Media, aka Deepfakes and its impact
46mTackling Deepfakes, Wars, and Other Security Threats in the GenAI World
16mThe Threat of Deepfakes to Remote Identity Verification Systems
27mIs now the time to solve the deepfake threat?
08mAdvancements in facial recognition