Should Australia's AI safety institute focus on future risks, present dangers or working with its international counterparts?