Lucius Bushnaq comments on CAIS-inspired approach towards safer and more interpretable AGIs

[Lucius Bushnaq]

Seems like a slight variant on MIRI’s visible thoughts project?