[ergonaught]

Putting aside AGI types of actual "intelligence", this is evolution at work, more or less. As soon as someone finds a useful means of automating the "fitness function" testing that the output has desirable qualities (correctness or entertainment or whatever), everything outside the extremes will rapidly leave human hands. The extremes fall soon thereafter.

The problems are mostly sociocultural, and I am reminded of "Agent Smith" commenting on humans rejecting a Matrix simulation that didn't suck for them.