Putting up Bumpers
Proposes an alignment strategy of layered audits, red-teaming, and post-deployment monitoring that catches and corrects misalignment in early AGI systems, even without solving alignment outright.
Discussion and research on AI alignment. alignmentforum.org
Proposes an alignment strategy of layered audits, red-teaming, and post-deployment monitoring that catches and corrects misalignment in early AGI systems, even without solving alignment outright.