The System · Part 1
Chapter 11: Generative AI and the End of Seeing Is Believing
Generative AI changes the production function of deception. The deeper threat isn't fakes, it's the 'liar's dividend': convincing forgeries undermine confidence in real evidence. Verification remains slow and expensive while generation is fast and cheap.
Key insights
- Detection accuracy for AI-generated content is declining
- The 'liar's dividend' means genuine evidence can be dismissed as fabricated
- Belief persists after correction, debunking rarely fully reverses perceptions
- Verification is slow and expensive; generation is fast and cheap
- Platform labeling hasn't altered engagement incentives amplifying synthetic content
- The 2024 election cycle tested the feared version and it did not materialize: AI content was a small fraction of election misinformation, simple manipulations dominated, and documented harms tracked the liar's dividend and fraud, confirming that the deepest risk is not persuasion by fake video but the erosion of confidence in evidence itself
The vulnerability itself is not new. The scale is.
Sections in this chapter
- Distortion to synthetic production
- Declining visual reliability
- The liar's dividend
- Speed and cost asymmetry
- Persistence of belief
- Public confidence trends