Why audio deepfakes are so hard to spot
Summary
Marketplace Tech host Stephanie Hughes interviews UC Berkeley professor Hani Farid about the rise of AI voice cloning and its potential for fraud. They discuss how difficult it is for humans to distinguish real voices from AI-generated ones, citing scams targeting individuals and enterprises. Farid advises using code words for verification and warns against sharing personal audio online, noting that while the tech has benefits, current deployment risks outweigh the positives.
Topics discussed
Sponsors and show introduction
Testing AI voice cloning with a live demo
Why AI voices are nearly indistinguishable from real ones
Real-world scams and fraud using voice cloning
Expert advice on protecting yourself from AI scams
Hani Farid's startup Get Real Security
Commercial break and sponsor messages
Risks of uploading personal data and photos of kids
Using code words to verify identity in calls
Ethical concerns and future of voice cloning tech
Outro and promotion for NYC Now
Listen ad-free on Castria