AI summary by TwinMind
AI’s Real Threat: Mass Manipulation
Sabine Hossenfelder
7m 34s
AI's Real Threat: Mass Manipulation Over Mass Unemployment
- AI's primary threat = mass manipulation, not mass unemployment
- Current AIs ideal for scammers due to personalization capabilities
- Used for political influence, product sales, decision manipulation
- Prevention difficult, but awareness helps avoid falling victim
Data Collection and Personal Profiling
- Data brokers collect extensive personal information
- Example broker claims: "most accurate identity" targeting real people vs devices/cookies
- US data: 7,000 individual attributes per person, 75% of purchase behavior tracked
- "Lola" example profile includes:
- Basic identity, viewing/reading habits, household members
- Social media follows, online/offline purchases
- Purchase timing, location, motivation ("why she buys")
Scientific Research on AI Persuasion
- University of Zurich study: 13 AI bots on Reddit's "change my view" forum
- 1,700+ comments posted
- Persuasion rate 6x higher than humans
- GPT-4 debate study: humans vs AI on topics (school uniforms, sugar taxes)
- With demographic information: AI won 64% of debates
- Term: "microtargeting" - future of marketing/political lobbying
- "Choice engineering with AI" research: using AI to engineer human decision-making
Real-World Implementation
- Google AI Max for search (May 2025)
- Automatically adjusts ads to match search queries
- Increases purchase rate by 14%
AI Scam Epidemic
- Celebrity impersonation scams:
- French woman lost €800,000 to fake Brad Pitt (voice cloning, fake photos)
- Similar Martin Henderson scam
- Dating profile scams: hyperrealistic profiles, personalized "lovebombing"
- US Federal Trade Commission (2024): 800,000+ impostor fraud reports, $3 billion losses
- Phishing evolution: AI-generated emails 54% click-through rate (matches human experts)
- Crypto scams: AI builds trust before theft, fake expert advice
- Elon Musk frequently deepfaked for crypto offers
- FBI (2024): 150,000 crypto complaints, $9+ billion losses
- Preceded Google V3 rollout
Reality Questioning and Trust Erosion
- Increasing skepticism about content authenticity
- Questions about posts, images, videos being real
- Personal note: speaker addresses avatar speculation due to lip issues
Data Protection Solutions
- Laws exist for data protection, removal requests possible
- Manual removal extremely time-consuming
- Incogn service automation:
- Contacts data brokers for removal requests
- Continuous monitoring and removal
- Progress updates provided
- 60% discount with code "Zabina"
