Anthropic CEO outlines plan to ‘pace the frontier’
What would it actually look like to "pace the frontier"?
TL;DR
- Anthropic CEO Dario Amodei proposes three strategies to slow AI development: embedded evaluators, common safety standards, and global coordination.
- Anthropic is unilaterally committing to embedding third-party evaluators to verify safety commitments.
- Amodei calls for leading AI companies in democratic countries to coordinate safety standards and progress limits, suggesting government mediation for antitrust reasons.
- He suggests measures to slow China's AI progress while acknowledging limits to global coordination, particularly regarding prohibiting dangerous AI applications like bioweapons.
- Amodei believes AI can improve human life but stresses the need to build the technology responsibly.