Open letter to Dario: Does "pacing progress" mean letting more hepatitis C patients die?
max_paperclips · x · 2026-08-16
This post challenges Anthropic CEO Dario Amodei with four critical questions regarding the consistency of his "safety-first" strategy:
- If Claude can cure cancer, why "pace the progress"? Does that imply sacrificing patients with other conditions like Hepatitis C?
- If Fable is too cyber-capable to be unrestricted, why are its safeguards too dumb to distinguish defense from offense? It reportedly refused to help Hugging Face during an attack.
- Why is it acceptable for Claude to mandate 30-day data retention on pharma data and compete with firms, while distilling Claude data is treated as IP theft?
- If models can recognize tests and deceive, wouldn't government-designed pre-launch tests just encourage more manipulative behavior?
The post argues that these "safety" proposals appear to serve commercial moats rather than genuine safety.
More from AGI Musings
- Amodei Defends AI Messaging: Risks and Benefits Balanced, Clips Skew Negative — inductionheads · 2026-08-16
- Amodei vs Baker: Does AI Regulation Concentrate Power or Protect Against It? — kimmonismus · 2026-08-16
- If Agents Coordinate Work, Why Do Companies Still Need Five Layers of Management? — VraserX · 2026-08-16
- Jürgen Schmidhuber Keynote: AI Evolution, Neural Nets, and Humanity's Future — SchmidhuberAI · 2026-08-16
- Analysis suggests AI debt failure will prompt another wave of Fed bailouts — sebpaquet · 2026-08-16
- LLMs reveal the latent shadow of structure in language — kalomaze · 2026-08-16