The Importance of Trustworthiness in AI

For AI to vastly surpass human performance at a task, there needs to be ample room for improvement above human level. How much headroom there is varies by task - and it's not clear how close humans are from the limits to competence in the particular domains that will matter. For instance, machines cannot win more often than the best humans at tic-tac-toe,. Or solve Rubik's cubes much more reliably, than extracting calories from fuel. So we are asking what we are asking exactly if AI can do much better than humans?