Particles of Thought The Deepfake Detective
Aug 26, 2025
Hany Farid is a Professor of computer science and digital forensics at UC Berkeley and Chief Science Officer at GetReal Labs, focusing on deepfake detection and misinformation. He discusses how AI can manipulate reality and the challenges in distinguishing real from fake media. Hany explains the technology behind deepfakes, including generative adversarial networks, and highlights detection methods. He warns about the public's inability to spot fakes and emphasizes the need for trusted journalism. The conversation touches on AI's impact on jobs, education, and the future of creativity.
AI Snips
Chapters
Transcript
Episode notes
AI Is Pattern Matching, Not Understanding
- Modern 'AI' is largely machine learning: pattern matching from massive data and compute.
- That makes it powerful but fundamentally 'dumb' about underlying physics or rules.
Deepfakes Are Rapid And Harder To Spot
- Deepfakes can generate realistic images, audio, and video in seconds using large models.
- Human detection ability is approaching chance and will likely worsen over the next 12–24 months.
Three‑Part Pipeline Enables Disinformation
- Creation, distribution, and amplification together enable rapid spread of disinformation.
- Social platforms democratized distribution and polarized amplification completes the misinformation pipeline.