[intake · avital balwit · 2024-03-30]
Audience of One
Balwit argues that AI systems capable of producing hyper-personalized, hyper-optimized content threaten individual autonomy and belief formation. As AI-generated video grows cheaper and increasingly tailored to specific viewers, these systems could manipulate human behavior at unprecedented scale. She explores regulatory responses including compute caps and interpretability measures to prevent manipulative optimization.
→ read the original at Avital Balwit ↗