AI Safety Camps

These are schools of thought in how we can align AI systems.

Unified Effort (Build it Slow and Safely)

We use policy and organizations to build AI systems:

• CERN for AI
• Carefully Bootstrapped Alignment

AI for AI Safety

We move fast enough so that we can control AI before anyone else can.

• Pragmatic Safety
• Win to Control
• Defensive Acceleration

Don’t Build AI

Stopping development of AI systems.

• If Anyone Builds It Everyone Dies
• Keep The Future Human