Peter N. Salib

If the foregoing arguments are right, then scenario three (risk apprehension, self-improvement, alignment) is the most likely of the six. This suggests a concrete near-term prediction: Contrary to standard arguments, when AIs achieve average human-le...