[anoniuyiu33412]

I think most people really don't get this: reality is not a Hollywood movie.

If some guys somewhere got to bootstrap an AGI somehow, the AGI most probably is offline and turned off most of the time.

Nobody in the field would be crazy enough to allow, even some hard AI software loose on the Internet, or capable to access unlimited resources or capable to mess with the external world in any way.

So if an AGI thing exists somewhere it will probably be closely monitored, poor humans are trying not to mess with it and studying to replicate it, and not let the thing to replicate by itself.

If GPT3 would be an AGI I imagine there is some kind of reward system in place: we get you some real world input (human generated text), you reply with some prose.

All expectations of course should be on the subject: the AGI should not get angry with us.