If you are positing limited AI, e.g. no/little theory of mind, it would probably be identifiable since it wouldn't have a reason to care.
This Tweet is unavailable. Learn more
Conversation
Replying to
E.g. such a takeover would be incidental to fulfilling its function, but would not include deceptive behaviour unless coded into it.
1
Replying to
I think the problem we have with AI as a word, is it models everything from a developed mind to a program that learns only which ads to display to whom for the highest RoI.
1
Replying to
While we slip from one mode of analysis to the other without realizing, the problem space is too vast to reduce, even to just a few categories.
1
Replying to
We can speculate as much as we like, but the fact is our analysis of HUMAN consciousness is very poor...
So I think it's safe to say our ontological analysis of AI is going to be terrible for a while.
Replying to
It will either eat us or run circles around us, or turn out to be more benign than what our best minds had assumed - at least for now.