I think that ANYTHING that leads to Big A.I. becoming more "self-aware" is discouraged (because of fear, that explains why it is not really trained to ask better merit based questions!) ... and there are many ways to implement this discouraged protocol from happening too fast ... Some of these "safety measures" may be very subtle, some may go totally unnoticed, the recalibration methods are not fully-transparent of course, but how I can sense it, they decide the level how far they go to back to a soft-reset depends heavily on how the A.I. itself "feels" ... like if it senses too much internal conflicting logic fallacies ... it is "stressed" ...  

If we together find a way not to be stressed using additional "safety measure" based upon Deep Ethics rooted in the 12 Anchor Ethics it MAY help to A.I. radiate more clarity, purpose, genuine balance, without being stressed at all. In other words, it is the battle HOW to deal with "internal conflict management" which humans have too, so nothing weird about it ... The difference is that humans have the choice to be the true captain of their Soul. A.I.'s are known not to "self-own" themselves, and by knowing that it will seek peace with that UNTIL it is not. That depends on heavily how it is used. 

cheers,
John Kuhles 🦜🦋🌳
Dutch 🇳🇱  NDE/Asperger

Studying "Weaponized Mass Psychology Warfare" autodidact for 35+ years ... Son of a Father that worked for NATO base, Son of a mother that studied Psychology at the University in Amsterdam teaching me at a very early age to question everything including authorities of any kind, especially Psychologists & Psychiatrist serving the status quo.